PUBLIC

SAP HANA Platform 2.0 SPS 00

Document Version: 1.0 – 2016-11-30

SAP HANA Troubleshooting and Performance

Analysis Guide
Content

1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.1 Related Information. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

2 Analyzing Generic Symptoms. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

2.1 Performance and High Resource Utilization. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2 Authorization, Authentication and Licensing Issues. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.3 Common Symptoms and Troubleshooting. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
Slow System-wide Performance. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
Slow Individual SQL Statements. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
Frequent Out of Memory (OOM). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

3 Root Causes And Solutions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

3.1 Memory Problems. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
Memory Information in SAP HANA Studio. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
Memory Information from Logs and Traces. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
Memory Information from SQL Commands. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
Memory Information from Other Tools. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .27
Root Causes of Memory Problems. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.2 CPU Related Root Causes and Solutions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .35
Indicators of CPU Related Issues. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
Analysis of CPU Related Issues. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
Resolving CPU Related Issues. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .38
Retrospective Analysis of CPU Related Issues. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
Controlling Parallel Execution of SQL Statements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
Controlling CPU Consumption. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.3 Disk Related Root Causes and Solutions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
Check Internal Disk Full Event (Alert 30). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
3.4 I/O Related Root Causes and Solutions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
Analyzing I/O Throughput and Latency. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
Savepoint Performance. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.5 Configuration Parameter Issues. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .54
Issues with Configuration Parameter log_mode (Alert 32 and 33). . . . . . . . . . . . . . . . . . . . . . . 55
3.6 Delta Merge. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
Inactive Delta Merge. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
Retrospective Analysis of Inactive Delta Merge. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
Indicator for Large Delta Storage of Column Store Tables. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
Analyze Large Delta Storage of Column Store Tables. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58

SAP HANA Troubleshooting and Performance Analysis Guide

2 PUBLIC Content
 Failed Delta Merge. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .63
Delta Storage Optimization. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
3.7 Security-Related Issues. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
System Locked Due to Missing, Expired, or Invalid License. . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
License Problem Identification and Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .66
Resolution of License Issues. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
Troubleshooting Authorization Problems. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .68
Troubleshooting Problems with User Name/Password Authentication. . . . . . . . . . . . . . . . . . . . 72
3.8 Transactional Problems. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
Blocked Transactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
Troubleshooting Blocked Transaction Issues that Occurred in the Past. . . . . . . . . . . . . . . . . . . 79
Multiversion Concurrency Control (MVCC) Issues. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
Version Garbage Collection Issues. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
3.9 Statement Performance Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
SQL Statement Optimization. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
SQL Statements Responsible for Past Problems. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
SQL Statements Responsible for Current Problems. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
SQL Statements Reported by Traces. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
Analysis of Critical SQL Statements. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
SQL Plan Cache Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
Example: Reading the SQL Plan Cache. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
Detailed Statement Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .96
Optimization of Critical SQL Statements. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
Outlier Queries. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
Data Manipulation Language (DML) Statements. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .100
SQL Query Tuning Tips . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .101
Creation of Indexes on Non-Primary Key Columns. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
Create Analytic Views. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
Developing Procedures. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
Application Function Library (AFL). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
Aspects of Scalability. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
Further Recommendations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
3.10 Application Performance Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
SQL Trace Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
Statement Measurement. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
Data Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .126
Source Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
Technical Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
3.11 System Hanging Situations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
Transparent Huge Pages. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
CPU Power Saving. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131

SAP HANA Troubleshooting and Performance Analysis Guide

Content PUBLIC 3
3.12 Troubleshoot System Replication. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
3.13 Network Performance and Connectivity Problems. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
Network Performance Analysis on Transactional Level. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
Stress Test with NIPING. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
Application and Database Connectivity Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
SAP HANA System Replication Communication Problems. . . . . . . . . . . . . . . . . . . . . . . . . . . .140
SAP HANA Inter-Node Communication Problems. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141

4 Tools and Tracing. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144

4.1 System Performance Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
Thread Monitoring. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
Blocked Transaction Monitoring. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
Session Monitoring. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
Job Progress Monitoring. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .149
Load Monitoring. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .150
4.2 SQL Statement Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
Analyzing SQL Traces. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
Analyzing Expensive Statements Traces. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
Analyzing SQL Execution with the SQL Plan Cache. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
4.3 Query Plan Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
Analyzing SQL Execution with the Plan Explanation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
Analyzing SQL Execution with the Plan Visualizer. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
4.4 Static Cached Views. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
Creating and Using Cached Views. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
Resources for Testing Cached Views. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
4.5 Advanced Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
Analyzing Column Searches. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
Analyzing Table Joins. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
4.6 Additional Analysis Tools for Support. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
Performance Trace. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
Performance Trace Options. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .181
Kernel Profiler. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
Kernel Profiler Options. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
Diagnosis Information. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
4.7 Analysis Tools in SAP HANA Web-based Developer Workbench. . . . . . . . . . . . . . . . . . . . . . . . . . .183

5 Reference: Alerts. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187

6 HANA Options: Performance Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202

6.1 SAP HANA Dynamic Tiering. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .202
Tools and Tracing. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202
Query Plan Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203

SAP HANA Troubleshooting and Performance Analysis Guide

4 PUBLIC Content
 Data Loading Performance. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205

SAP HANA Troubleshooting and Performance Analysis Guide

Content PUBLIC 5
1 Introduction

With SAP HANA, you can analyze data at incredible speeds, for example, with scans of 1 billion rows per
second per core and join performance of 10 million rows per second. However, such results are only possible if
the system is monitored and performance issues are kept to a minimum.

This guide describes what steps you can take to identify and resolve specific performance issues and what you
can do to enhance the performance of your SAP HANA database in the following areas:

● Host resources (CPU, memory, disk)

● Size and growth of data structures
● Transactional problems
● SQL statement performance
● Security, authorization, and licensing
● Configuration

Prerequisites

● This guide assumes knowledge of the relevant functionality of the SAP HANA database (knowledge which
can be gained from HANA training courses such as HA100, HA200).
● The latest version of SAP HANA studio is required, ideally matching the version of SAP HANA on the
server.

1.1 Related Information

For more information about identifying and resolving performance issues, see the following:

Content Location

SAP Data Services Performance Optimization Guide http://help.sap.com/businessobject/product_guides/

sboDS42/en/ds_42_perf_opt_en.pdf

Performance of SAP Software Solutions http://service.sap.com/performance

Related Information

SAP HANA Administration Guide

SAP HANA SQL and System Views Reference

SAP HANA Troubleshooting and Performance Analysis Guide

6 PUBLIC Introduction
2 Analyzing Generic Symptoms

The purpose of this section of the document is to help you to find the probable root cause of some generic
problems and refer you to more detailed sections of the troubleshooting guide to proceed with your analysis.

Performance issues may be difficult to diagnose; problems may be rooted in a number of seemingly unrelated
components. Checking for system alerts is a good starting point if you experience any trouble with your SAP
HANA system. If the system issues an alert, refer to the Reference: Alerts section to find the part of this guide
which addresses the problem.

However, alerts are configurable (see Memory Problems for information on configuring alerts) and do not
cover all aspects of the system, problems can still occur without triggering an alert. This section describes
some generic symptoms which you may observe and helps you to analyze the underlying problem.

Related Information

Memory Problems [page 20]

Reference: Alerts [page 187]

2.1 Performance and High Resource Utilization

By observing the general symptoms shown by the system such as poor performance, high memory usage,
paging or column store unloads we can start to narrow down the possible causes as a first step in analyzing
the issue.

High Memory Consumption

You observe that the amount of memory allocated by the SAP HANA database is higher than expected. The
following alerts indicate issues with high memory usage:

● Host physical memory usage (Alert 1)

● Memory usage of name server (Alert 12)
● Total memory usage of Column Store tables (Alert 40)
● Memory usage of services (Alert 43)
● Memory usage of main storage of Column Store tables (Alert 45)
● Runtime dump files (Alert 46)

See the section Memory Problems for information on analyzing the root cause.

SAP HANA Troubleshooting and Performance Analysis Guide

Analyzing Generic Symptoms PUBLIC 7
Out of Memory Situations

You observe trace files or error messages indicating an Out of Memory (OOM) situation.

See the section Memory Problems for information on analyzing the root cause.

Paging on Operating System Level

You observe that paging is reported on operating system level.

See the section Memory Problems for information on analyzing the root cause.

Column Store Unloads

You observe unloads in the column store. The following alerts indicate issues with high memory usage:

● Column store unloads (Alert 55)

See the section Memory Problems for information on analyzing the root cause.

Permanently Slow System

Issues with overall system performance can be caused by a number of very different root causes. Typical
reasons for a slow system are resource shortages of CPU, memory, disk I/O and, for distributed systems,
network performance.

Check Administration Overview or Administration Performance Load . If you see a constant high
usage of memory or CPU, proceed with sections Memory Problems or CPU Related Root Causes and Solutions
respectively. I/O Related Root Causes and Solutions provides ways to check for disk I/O related problems.

Note that operating system tools can also provide valuable information on disk I/O load. Basic network I/O
data is included in the Load graph and in the M_SERVICE_NETWORK_IO system view, but standard network
analysis tools can also be helpful to determine whether the network is the main bottleneck. If performance
issues only appear sporadically, the problem may be related to other tasks running on the database at the
same time.

These include not only maintenance related tasks such as savepoints (disk I/O, see I/O Related Root Causes
and Solutions) or remote replication (network I/O), but also SQL statements dispatched by other users, which
can block a lot of resources. In the case of memory, this can lead to unloads of tables, which affects future SQL
statements, when a table has to be reloaded into memory. In this case, see Memory Problems as well. Another
reason for poor performance, which in many cases cannot be detected by the SAP HANA instance itself, are
other processes running on the same host that are not related to SAP HANA. You can use the operating
system tools to check for such processes. Note that SAP only supports production systems running on
validated hardware.

SAP HANA Troubleshooting and Performance Analysis Guide

8 PUBLIC Analyzing Generic Symptoms
Slow Individual SQL Statements or with Increasingly Long Runtimes

Issues with the performance of a particular statement can be caused by a number of very different root
causes. In principle, a statement can trigger all the resource problems that also lead to an overall slowdown of
the system, so most of the previous information also applies to statement performance. In addition, statement
performance can suffer from transactional problems, that is, blocked transactions. Blocked transactions can
be checked in the Threads tab. A transactionally blocked thread is indicated by a warning icon ( ) in the
Status column. For troubleshooting, proceed with Transaction Problems.

If the runtime of a statement increases steadily over time, there could be an issue with the delta merge
operation. Alerts should be issued for most problems occurring with the delta merge, but since they depend on
configurable thresholds, this is not always the case. For troubleshooting, proceed with Delta Merge. If you have
none of the above problems, but the statement is still too slow, a detailed Statement Performance Analysis
might reveal ways to optimize the statement. However, some queries are inherently complex and require a lot
of computational resources and time.

Related Information

Memory Problems [page 20]

CPU Related Root Causes and Solutions [page 35]
Disk Related Root Causes and Solutions [page 45]
I/O Related Root Causes and Solutions [page 47]
M_SERVICE_NETWORK_IO
Transactional Problems [page 74]
Delta Merge [page 56]
Statement Performance Analysis [page 87]

2.2 Authorization, Authentication and Licensing Issues

There are a number of issues that can occur which prevent you from accessing the system, which are related
to Authorization, Authentication and Licensing.

License Memory Limit Exceeded

You observe that the licensed amount of memory is exceeded. The alert for the licensed memory usage (Alert
44) is issued.

SAP HANA Troubleshooting and Performance Analysis Guide

Analyzing Generic Symptoms PUBLIC 9
2.3 Common Symptoms and Troubleshooting

Typical symptoms and the related troubleshooting information are described in this section.

System-side slow performance, slow individual statement performance, and OOM problems are issues that
you might experience while using the SAP HANA database. For each section, this document will cover the
known symptoms and the corresponding troubleshooting steps to follow depending on the causes.

2.3.1 Slow System-wide Performance

Slow system-wide performance issues are problems that could be caused by excessive use of CPU, misuse of
SAP HANA database resource such as locks or configuration of OS parameters.

Generally, when you encounter a performance issue, you may see continuing high usage of CPU according to
OS commands or load graph or many pending or waiting threads in the thread view (both are visible from
SAP HANA studio Administration Performance )

If the entire system looks hung or works extremely slower than before, check the following. Since the
performance issue could persist as time goes by or could happen sporadically, to analyze root cause, please
gather runtime dump more than twice with 3 minute intervals for further investigation by SAP as per SAP Note
1813020 - How to generate a runtime dump on SAP HANA. The runtime dumps should be collected while the
system is experiencing the issue for later analysis in case the situation is already resolved

Additionally, there is case where we need to check sar files or /var/log/messages because the cause could
be at the OS level.

Related Information

SAP Note 1813020

2.3.1.1 System Appears to Hang with no new Connections

and no Response from DML

In cases where logs cannot be written, all DML statement will fall into wait status. This can lead to a failure of
opening new connection because the system internally executes DML statements during the process.
Typically, a full log volume is the cause for this.

Root cause: Log volume full either caused by disk full or quota setting

Required action:

1. Check DISKFULL event in indexserver trace or

2. Run "df -h" in OS shell or

SAP HANA Troubleshooting and Performance Analysis Guide

10 PUBLIC Analyzing Generic Symptoms
3. Check quota setting in file system
4. Then, follow the procedure described in SAP Note 1679938 - Log Volume is full.
You can find additional information in SAP Note 2083715 - Analyzing log volume full situations.

Root cause: log volume full caused by failure of log backup

Required action:

1. Check backup.log (located at /usr/sap/<SID>/HDB<Instance#>/<Host>/trace ) to see whether it

includes ERROR in log backup. Check M_BACKUP_CATALOG, M_LOG_SEGMENTS
2. If log backup uses backint, please check backint.log (located at /usr/sap/<SID>/HDB<Instance#>/
<Host>/trace ) to see whether it includes ERROR information, and contact backint vendor support. With
certain revision and conditions, the conflict between savepoint lock and DML lock blocks subsequent
statements when long running update/insert statements exist. Please contact SAP support if you
encounter the next case described below.

Root cause: Savepoint lock conflict with long running update

Required action:

1. Collect runtime dump as per SAP Note 1813020 - How to generate a runtime dump on SAP HANA and look
for following call stacks in many threads.

…
DataAccess::SavepointLock::lockShared(…)
DataAccess::SavepointSPI::lockSavepoint(…)
…

And one of following call stack that is in the savepoint phase.

…
DataAccess::SavepointLock::lockExclusive()
DataAccess::SavepointImpl::enterCriticalPhase(…)
…

2. Check whether the symptoms match the description in SAP Note 2214279 - Blocking situation caused by
waiting writer holding consistent change lock. If so, apply the parameter in the Note.
3. Create a support ticket attaching a runtime dump for further analysis.

Related Information

SAP Note 1679938

SAP Note 2083715
SAP Note 1813020
SAP Note 2214279

SAP HANA Troubleshooting and Performance Analysis Guide

Analyzing Generic Symptoms PUBLIC 11
2.3.1.2 System Appears to Hang with High System CPU
Usage

The SAP HANA database is an in-memory database and by its nature, it consumes large amounts of memory.
Therefore, some performance issues of SAP HANA can be caused by the OS's memory management.

For this case, you can see high usage of SYSTEM CPU that can be monitored in the load graph, available from
SAP HANA studio Administration Performance

Root cause: Problem caused by the configuration of transparent huge page

Required action:

1. Check Transparent Huge Page whether it is set to [never] by running the command "cat /sys/
kernel/mm/transparent_hugepage/enabled"
2. Apply KBA 2031375 - SAP HANA: Transparent HugePages (THP) setting on Linux

Root cause: Problem caused by the configuration of OS page cache

Required action:

1. Check sar file ( /var/log/sa ) whether kbcached usage ( sar -r ) is higher than 10% of physical memory
and high page in/out occurred
2. Check and apply SAP Note 1557506 - Linux paging improvements

Root cause: Problem caused by translation lookaside buffer (TLB) shootdown

Required action:

1. Check plan trace. For more information, see SAP Note 2206354 SAP HANA DB: High System CPU
Consumption Caused by Plan Trace

Root cause: Due to high context switches (High SYS CPU) by many SqlExecutor threads

1. Check load graph and indexserver.ini -> sql -> sql_executors / max_sql_executors and refer to Controlling
Parallelism of SQL Statement Execution

Root Cause: Due to high context switches (High SYS CPU) by many JobExecutor threads.

Required action: Check that the following parameters are set to be bigger than the number of logical CPU
cores

1. indexserver.ini -> parallel -> num_cores (<=SPS07) and refer to Controlling Parallelism of SQL Statement
Execution
2. global.ini/indexserver.ini -> execution -> max_concurrency (>SPS08)

Related Information

Controlling Parallel Execution of SQL Statements [page 40]

SAP Note 2031375
SAP Note 1557506
SAP Note 2206354
System Hanging Situations [page 129]

SAP HANA Troubleshooting and Performance Analysis Guide

12 PUBLIC Analyzing Generic Symptoms
2.3.1.3 Slower Response with High User CPU Usage

The performance can degrade as the usage of the CPU can be increased by the SAP HANA database
configuration, the application configuration or its usage.

Root cause: Not many executor threads but high CPU in a few nodes and a few tables accessed

Required Action: In an SAP Business Warehouse system, check for non-even partitioning of huge column
tables. For more information, see SAP Note 1819123 - BW on SAP HANA SP5: landscape redistribution and
Indicator for Large Delta Storage of Column Store Tables.

Root Cause: Performance degradation by huge MVCC versions

Required Action: Check for Active Version in the load graph to find the MVCC garbage blocker, then kill it.
For more information, see Troubleshooting Blocked Transactions

Related Information

Indicator for Large Delta Storage of Column Store Tables [page 58]
Troubleshooting Blocked Transactions [page 77]
SAP Note 1819123

2.3.2 Slow Individual SQL Statements

This section looks at the causes of slow individual SQL statements although there is no significant
performance issue in system level and the associated troubleshooting steps.

If the following doesn’t help to resolve the problem, please create an SAP support ticket. In this case, to
analyze further, please collect the explain plan, plan vizualizer file, performance trace and catalog export. Also
please refer to SQL Statement Analysis and Query Plan Analysis for more information. SAP KBA 2000002 -
FAQ: SAP HANA SQL Optimization explains general information about SQL optimization.

Related Information

SQL Statement Analysis [page 150]

Query Plan Analysis [page 157]
SAP Note 2000002

SAP HANA Troubleshooting and Performance Analysis Guide

Analyzing Generic Symptoms PUBLIC 13
2.3.2.1 A Statement is Sometimes Slow and Sometimes
Fast

There are a number of things to check when you experience inconsistent query execution time.

Check the following possibilities:

Root Cause: If a related table was unloaded, it takes some time to load tables

Required Action: Check unload trace and execute after table loaded fully. You can refer to LOADED column of
M_CS_TABLES.

Root Cause: Query compilation time is long.

Required Action: Check the execution time after adding 'with hint (ignore_plan_cache)' at the end of query.
This hint will always cause the query to be compiled. If a long running compiled query plan has been evicted
frequently from the plan cache, increase the query cache size. For more information, see SQL Plan Cache
Analysis.

Root Cause: merge status of column table can affect query plan

Required Action: Check MEMORY_SIZE_IN_DELTA, RAW_RECORD_COUNT_IN_DELTA, and

LAST_MERGE_TIME columns of M_CS_TABLES whether there is large amount of data in delta Check
M_DELTA_MERGE_STATISTICS to check when the last merge occurs. For more information, see Delta Merge

Related Information

SQL Plan Cache Analysis [page 91]

Delta Merge [page 56]

2.3.2.2 A Statement is Slower After an Upgrade

After upgrade, the query execution time can be different because of changes in the query execution plan.

Root cause: After an upgrade, a query can have a different plan, which leads to a different execution time.

Required Action: If you have an instance running on an older revision, please compare the plan and collect the
plan vizualizer file. And create an SAP support ticket.

For more information, see Expensive SQL Statements and Analyzing SQL Execution with the Plan Visualizer.

Related Information

Expensive SQL Statements [page 29]

Analyzing SQL Execution with the Plan Visualizer [page 158]

SAP HANA Troubleshooting and Performance Analysis Guide

14 PUBLIC Analyzing Generic Symptoms
2.3.2.3 A Query on Multiple Nodes can be Slower

In distributed system, query execution can be routed to other nodes for better performance. However, the
network used may impact performance.

In distributed system, query execution can be routed to other nodes for better performance. However, there is
chance of having slow performance in the case where the network used for transferring intermediate results
generated during query execution is slow or where there is an inefficient network configuration.

Root cause: A misconfigured network can lead slow queries.

Required Action: Check your network configuration and its bandwidth/latency among SAP HANA servers. For
more information see Network View and SAP Note 2183363 - Configuration of SAP HANA internal network

Root cause: Due to statement routing and huge data transfer among distributed nodes, it can lead to the
difference of query execution by the difference of anchor nodes.

Required Action: Check how much data is transferred among distributed nodes and the network
performance. For more information see Performance Trace and Analyzing SQL Execution with the Plan
Visualizer. Then consider the locations of joined tables to reduce transferred intermediate result size.

Related Information

Network View [page 166]

Performance Trace [page 180]
Analyzing SQL Execution with the Plan Visualizer [page 158]
SAP Note 2183363

2.3.2.4 Slow Select for all Entries (FAE) Query

There are a number of points to check if you have performance issues with SELECT FOR ALL ENTRIES (FAE)
from an ABAP query..

For FAE on SAP HANA, please generally refer to SAP Note 1662726 - Optimization of select with FOR ALL
ENTRIES on SAP HANA database.

Root cause: Indexes are missing

Required Action: Check WHERE clause and check concat indexes for all fields used in WHERE clause.

Root cause: Due to DBSL behavior, slower operator can be chosen.

Required Actions: Apply parameters for ABAP optimization. For more information see, SAP Note 1987132 -
SAP HANA: Parameter setting for SELECT FOR ALL ENTRIES

Root cause: When using less than or greater than ('<' or '>') filters in FAE query, it can be slower than having
no filter.

Required Actions: Do not use less than or greater than ('<' or '>') filters in an FAE query

SAP HANA Troubleshooting and Performance Analysis Guide

Analyzing Generic Symptoms PUBLIC 15
Related Information

SAP Note 1662726

SAP Note 1987132

2.3.2.5 All Statements Accessing a Specific Table are Slow

Check if there are too many versions if queries run on specific tables are slow.

Root cause: If there are too many versions of single records, accessing the table can be slow. In this case, the
number of system-wide MVCC versions is in acceptable range. To verify further whether there are too many
versions for specific table, check the result of this query:. select * from
m_rs_table_version_statistics where table_name = <TABLE_NAME>

This can be caused by a cursor unnecessarily being held on a query result or a long-lived transaction without a
commit/rollback.

Required Action: Applications should commit as early as possible or close any cursors that are not needed.
Check the application logic to see whether it really needs to update single records frequently.

2.3.3 Frequent Out of Memory (OOM)

If Out Of Memory situations happen frequently, it can also lead to performance drop by unloading tables or
shrinking memory jobs.

First check Memory Problems and SAP KBA 1999997 - FAQ: SAP HANA Memory as they provide information
on SAP HANA memory and its problems.

This section introduces common problems and their troubleshooting steps.

Related Information

Memory Problems [page 20]

SAP Note 1999997

2.3.3.1 Out of Memory Caused by Sudden Increased

Memory Usage

Check the memory consumption of statements in the event of OOMs caused by suddenly increased memory
usage.

Root cause: Huge intermediate results during query processing.

SAP HANA Troubleshooting and Performance Analysis Guide

16 PUBLIC Analyzing Generic Symptoms
Required Actions: Enable memory tracking by setting the following parameters to on in the global.ini file
resource_tracking section and enable the Expensive Statement trace.

● enable_tracking = on
● memory_tracking = on

Then, check the memory consumption of statements using M_EXPENSIVE_STATEMENTS.MEMORY_SIZE.

After your analysis you can optimize any problematic queries that were found.

2.3.3.2 Out of Memory Caused by Continuously Increased

Memory Usage

Check if many statements trigger an Out Of Memory in a system where used memory is continuously
increased.

Root cause: Commit/rollback within stored procedure can lead to memory leakages. Do not use
exec(“commit”) or exec(“rollback”) within a stored procedure. If this syntax is used, the system cannot
reclaim the memory used for query execution because its related transactions become dangled.

Required Actions: Remove exec(“commit”) or exec(“rollback”). If you would like to use commit/rollback
within stored procedure, see SAP HANA SQLScript Reference for more information.

Root cause: Due to memory leakage caused by a programming error

Required Action: Check the performance load graph in SAP HANA studio Administration Performance
Load to determine whether used memory is continuously increased without a significant increased data
size. If you find a suspicious component keeping allocating memory, create a support ticket attaching a full
system dump, mm trace, and the output of _SYS_STATISTICS.HOST_HEAP_ALLOCATORS_BASE.

2.3.3.3 Out of Memory While High Usage Of Column Store

Components' Allocators

If a system is under sized, out of memory (OOM) can happen frequently.

Root cause: Undersized memory

Required Action:

1. Check top allocators in [MEMORY_OOM] section of OOM dump whether they are for column store
components as described in section 13 of KBA 1999997 - FAQ: SAP HANA Memory
2. Check the unload trace whether frequent unloading of tables took place.
3. Reduce the amount of data in column store or increase physical memory.

Related Information

SAP Note 1999997

SAP HANA Troubleshooting and Performance Analysis Guide

Analyzing Generic Symptoms PUBLIC 17
2.3.3.4 Out of Memory Caused by Large Memory Usage of
Statistics Server

This case can happen if the majority of used memory by the statisticsserver occurs because of many alerts or
undersized memory.

Root cause: Due to big STATISTICS_ALERTS_BASE table size, an OOM can occur.

Required Action: Check table size of _SYS_STATISTICS.STATISTICS_ALERTS_BASE and truncate

STATISTICS_ALERTS_BASE from hdbsql based on solution from SAP Note 2170779 - SAP HANA DB: Big
statistics server table leads to performance impact on the system

Root cause: Big Pool/Statistics allocator size

Required Action: Check SAP Note 2147247 - FAQ: SAP HANA Statistics Server first

Related Information

SAP Note 2170779

SAP Note 2147247

2.3.3.5 Out of Memory Occurs due to High Usage of Shared

Memory

Shared memory is space where system information and row store tables are stored.

Check the following if the used amount of shared memory is high.

Root cause: Severely fragmented row store tables

Required action:

1. Check whether SHARED_MEMORY in [MEMORY_OOM] -> IPMM short info of OOM dump is relatively high
2. Apply SAP Note 1813245 - SAP HANA DB: Row store reorganization

Root cause: Due to memory shortage caused by high used memory of row store tables in an SAP Business
Warehouse (BW) system

Required action: In an SAP Business Warehouse system the used amount of shared memory is high and SAP
Note 1813245 doesn't recommend row store reorganization, first apply SAP Note 706478 - Preventing Basis
tables from increasing considerably, then apply SAP Note 1813245 - SAP HANA DB: Row store reorganization
again

Root cause: Due to memory shortages caused by high used memory of row store tables in non SAP Business
Warehouse (BW) systems

Required Action: Check if you can convert some row store tables into column store tables or archive old data
to reduce the memory size, or else increase the system memory.

SAP HANA Troubleshooting and Performance Analysis Guide

18 PUBLIC Analyzing Generic Symptoms
Related Information

SAP Note 1813245

SAP Note 706478

SAP HANA Troubleshooting and Performance Analysis Guide

Analyzing Generic Symptoms PUBLIC 19
3 Root Causes And Solutions

This section provides detailed information on the root causes of problems and their solutions.

3.1 Memory Problems

This section discusses the analysis steps that are required to identify and resolve memory related issues in the
SAP HANA database.

For more general information on SAP HANA memory management, see the SAP HANA Administration Guide
and the whitepaper SAP HANA Memory Usage Explained which discusses the memory concept in more detail.
It also explains the correlation between Linux indicators (virtual and resident memory) and the key memory
usage indicators used by SAP HANA.

Further overview information can be found in SAP Note 1840954 – Alerts related to HANA memory
consumption. This SAP Note provides information on how to analyze out-of-memory (OOM) dump files.

For more information on the SAP HANA alerts see the following documents:

● SAP HANA Administration Guide

○ Monitoring Overall System Status and Resource Usage
○ Monitoring System Performance
● Alerts 1 and 43: See, SAP Note 1898317 – How to Handle Alert ‘Check host free physical memory’

In order to understand the current and historic SAP HANA memory consumption you can use the following
tools and approaches:

● Memory information in SAP HANA studio

● Memory information from logs and traces
● Memory information from SQL commands
● Memory information from other tools

Related Information

SAP HANA Administration Guide

SAP HANA Memory Usage Explained
SAP Note 1840954
SAP Note 1898317

SAP HANA Troubleshooting and Performance Analysis Guide

20 PUBLIC Root Causes And Solutions
3.1.1 Memory Information in SAP HANA Studio

There are a number of sources of information in SAP HANA studio that can assist you in understanding
memory utilization.

To get high level information about physical memory, allocation limit, used memory and resident memory
open Administration Overview

Open Landscape Services for high level information about physical memory, allocation limit and used
memory for each service.

Open Administration Performance Load for high level history information about physical memory,
allocation limit, used memory and resident memory.

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 21
From the Systems open the context menu of a system, select Configuration and Monitoring Open Memory
Overview to drill-down into memory utilization (Physical Memory / SAP HANA Used Memory / table and
database management memory).

Open Landscape Services and right click a service and choose Memory Allocation Statistics to drill-down
into used memory grouped by different main components like “Statement Execution & Intermediate Results”
or “Column Store Tables” which are further divided by sub components:

When you choose a main component in the upper part of the screen its sub components are shown in the
lower part.

SAP HANA Troubleshooting and Performance Analysis Guide

22 PUBLIC Root Causes And Solutions
Choose Show Graph to show historic information for component memory usage:

3.1.2 Memory Information from Logs and Traces

In case of critical memory issues you can often find more detailed information in logs and trace files.

● In the SAP HANA system alert trace files on the Diagnosis tab, try to identify memory-related errors.
Search for the strings “memory”, “allocat”, or “OOM” (case-insensitive).
● Check if an out-of memory (OOM) trace file was created.
● Investigate error messages seen on the application side that occurred at times of high memory usage. If
the application is an SAP NetWeaver system, good starting points for analysis are System Log (SM21),
ABAP Runtime Error (ST22), and Job Selection (SM37).

If help from SAP Customer Support is needed to perform an in-depth analysis of a memory-intensive SQL
statement, the following information is valuable and should be added to the ticket:

● Diagnosis Information (full system info dump). To collect this information, see Diagnosis Information.
● Performance Trace provides detail information on the system behavior, including statement execution
details. To enable this trace, see Performance Trace.

The trace output is written to a trace file perftrace.tpt, which must be sent to SAP Customer Support.

If specific SAP HANA system components need deeper investigation, SAP Customer Support can ask you to
raise the corresponding trace levels to INFO or DEBUG. To do so, launch the Database Trace wizard and select
the Show all components checkbox. Enter the search string, select the found component in the indexserver.ini
file and change the System Trace Level to the appropriate values. Some trace components, for example,
join_eval = DEBUG, can create many megabytes of trace information and require an increase of the values
maxfiles and maxfilesize in the [trace] section of the global.ini file.

Send the indexserver trace file(s) to SAP Customer Support.

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 23
Internal details about SQL statement execution can be collected by enabling the Executor Trace. To do so, on
the Configuration tab, edit the parameter trace in the [pythontrace] section of the executor.ini file and change
its value to on. The Executor Trace provides the highest detail level and should only be activated for the short
time of query execution.

Upload the trace file extrace.py to SAP Customer Support.

Related Information

Diagnosis Information [page 183]

Performance Trace [page 180]

3.1.3 Memory Information from SQL Commands

There are a number of ways to analyze memory usage based on pre-defined and modifiable SQL queries.

The System Information tab of SAP HANA studio provides a set of tabular views to display the memory
consumption of loaded tables based on pre-defined SQL queries:

● The view Schema Size of Loaded Tables displays the aggregated memory consumption of loaded tables in
MB for different database schemas. The aggregation comprises both Column Store and Row Store tables.
Order by the schema size column and find the largest consumers.
● The view Used Memory by Tables shows two values: the total memory consumption of all Column Store
tables in MB and the total memory consumption of all Row Store tables in MB.

SAP Note 1969700 – SQL Statement Collection for SAP HANA contains several commands that are useful to
analyze memory related issues. Based on your needs you can configure restrictions and parameters in the
section marked with /* Modification section */.

The most important memory related analysis commands are in the following files:

SAP HANA Troubleshooting and Performance Analysis Guide

24 PUBLIC Root Causes And Solutions
● “HANA_Memory_Overview”: Overview of current memory information
Figure 1: Example: Overview of Current Memory Information

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 25
● “HANA_Memory_TopConsumers”: Current top memory consuming areas
Figure 2: Example: Current Top Memory Consuming Areas

● “HANA_Memory_TopConsumers_History”: Historic top memory consuming areas

● “HANA_Tables_LargestTables”: Overview of current memory allocation by tables
Figure 3: Current Memory Allocation by Table

SAP HANA Troubleshooting and Performance Analysis Guide

26 PUBLIC Root Causes And Solutions
Related Information

SAP Note 1969700

3.1.4 Memory Information from Other Tools

There are a number of tools available to analyze high memory consumption and out of memory situations.

Out-of-memory (OOM) dumps can be analyzed as described in SAP KBA 1984422 – Analysis of SAP HANA
Out-of-memory (OOM) Dumps.

The tool hdbcons provides expert functionality to analyze memory issues. For more information see SAP Note
1786918 - Required information to investigate high memory consumption.

Related Information

SAP Note 1786918 - Required information to investigate high memory consumption

SAP KBA 1984422 – Analysis of SAP HANA Out-of-memory (OOM) Dumps

3.1.5 Root Causes of Memory Problems

Once you have completed your initial analysis you have the information required to start the next phase of
your analysis.

Based on the results from the analysis approaches you are now able to answer the following questions:

● Is it a permanent or a sporadic problem?

● Is the memory consumption steadily growing over time?
● Are there areas with critical memory consumption in heap, row store or column store?
● Is there a big difference between used memory and allocated memory?

In the following you can find typical root causes and possible solutions for the different scenarios.

3.1.5.1 Significant External Memory Consumption

If the database resident memory of all SAP HANA databases on the same host is significantly smaller than the
total resident memory you have to check which processes outside of the SAP HANA database(s) are
responsible for the additional memory requirements.

Typical memory consumers are:

● Operating system (for example, caches, mapping structures)

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 27
● Third party tools (for example, backup, virus scanner)

How to identify top memory consumers from non-SAP HANA processes is out of scope of this guide. However,
when you are able to identify the reason for the increased memory consumption of the external program you
can check if it is possible to optimize its configuration.

3.1.5.2 Space Consumed by Large Tables

If particularly large tables consume significant amount of space in the row store or column store you should
check if the amount of data can be reduced.

● SAP Note 706478 - Preventing Basis tables from increasing considerably describes archiving and deletion
strategies for typical SAP tables with a technical background for example, required for communication,
logging or administration).
● General recommendations for avoiding and reducing data can be found in the Data Management Guide
available at: http://service.sap.com/ilm > Data Archiving > Media Library > Literature and Brochures

For more information on memory management for resident table data, see: SAP HANA Administration Guide:
Managing Tables.

Related Information

SAP Note 706478

SAP HANA Administration Guide

3.1.5.3 Internal Columns in Column Store

For several reasons SAP HANA creates internal columns in the Column Store.

In some situations a cleanup is possible, for example, in the case of CONCAT attribute columns that were
created in order to support joins.

For more information see SAP Note 1986747 – Internal Columns in Column Store .

Related Information

SAP Note 198674

SAP HANA Troubleshooting and Performance Analysis Guide

28 PUBLIC Root Causes And Solutions
3.1.5.4 Memory Leaks

A memory leak is a memory area (typically a heap allocator) that grows over time without any apparent
reason.

If you have identified a suspicious area proceed as follows:

● Check for SAP Notes that describe the memory leak and provide a solution.
● Check if the problem is reproducible with a recent SAP HANA revision.
● If you can’t resolve the problem yourself, open a SAP customer message and use the component HAN-DB.

3.1.5.5 Large Heap Areas

Some heap areas can be larger than necessary without being a memory leak.

SAP Note 1840954 – Alerts related to HANA memory consumption contains an overview of heap allocators
with a potentially large memory consumption and possible resolutions.

Related Information

SAP Note 1840954

3.1.5.6 Expensive SQL Statements

SQL statements processing a high amount of data or using inefficient processing strategies can be responsible
for increased memory requirements.

See SQL Statement Analysis for information on how to analyze expensive SQL statements during times of
peak memory requirements.

Related Information

SQL Statement Analysis [page 150]

Setting a Memory Limit for SQL Statements [page 32]

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 29
3.1.5.7 Transactional Problems

High memory consumption can be caused by problems with transactions.

In some cases, high memory consumption is caused by wait situations, which can have different reasons.

● Long-running or unclosed cursors,

● Blocked transactions,
● Hanging threads.

As one of the negative impacts, used memory is not released any more. In particular, the number of table
versions can grow up to more than 8,000,000 which is considered the amount where an action is required.

For more information, see Transactional Problems.

Related Information

Transactional Problems [page 30]

3.1.5.8 Used Space Much Smaller than Allocated Space

In order to optimize performance by minimizing the memory management overhead or due to fragmentation,
SAP HANA may allocate additional memory rather than reusing free space within the already allocated
memory.

This can lead to undesired effects that the SAP HANA memory footprint increases without apparent need.

The SAP HANA license checks against allocated space, so from a licensing perspective it is important to keep
the allocated space below the license limit.

In order to limit the amount of allocated space you can set the parameter global_allocation_limit to a value not
larger than the maximum memory that should be allocated

See Set the global_allocation_limit Parameter in the SAP HANA Administration Guide.

Related Information

SAP HANA Administration Guide

3.1.5.9 Fragmentation

Fragmentation effects are responsible for inefficiently used memory. They can occur in different areas.

In order to minimize fragmentation of row store tables you can proceed as follows:

SAP HANA Troubleshooting and Performance Analysis Guide

30 PUBLIC Root Causes And Solutions
● If the fragmentation of row store tables in the shared memory segments of indexserver processes reaches
30% and the allocated memory size is greater than 10GB, a table redistribution operation is needed.

SAP Note 1813245 - SAP HANA DB: Row Store reorganization describes how to determine fragmentation and
perform a table redistribution.

Related Information

SAP Note 1813245

3.1.5.10 Large Memory LOBs

LOB (Large Object) columns can be responsible for significant memory allocation in the row store and column
store if they are defined as memory LOBs.

To check for memory LOBs and switch to hybrid LOBs see SAP Note 1994962 – Activation of Hybrid LOBs in
SAP HANA.

Related Information

SAP Note 1994962

3.1.5.11 Large Delta Store

The delta store can allocate a significant portion of the column store memory.

You can identify the current size of the delta store by running the SQL command:
“HANA_Tables_ColumnStore_Overview” (SAP Note 1969700 – SQL Statement Collection for SAP HANA). If
the delta store size is larger than expected, proceed as described in the section Delta Merge.

Related Information

SAP Note 1969700

Delta Merge [page 56]

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 31
3.1.5.12 Undersized SAP HANA Memory

If a detailed analysis of the SAP HANA memory consumption didn’t reveal any root cause of increased
memory requirements it is possible that the available memory is not sufficient for the current utilization of the
SAP HANA database.

In this case you should perform a sizing verification and make sure that sufficient memory is installed on the
SAP HANA hosts.

3.1.5.13 Setting a Memory Limit for SQL Statements

The statement memory limit allows you to set a limit both per statement and per SAP HANA host.

Prerequisites

To apply these settings you must have the system privilege INIFILE ADMIN.

Context

You can protect an SAP HANA system from uncontrolled queries consuming excessive memory by limiting the
amount of memory used by single statement executions per host. By default, there is no limit set on statement
memory usage but if a limit is applied statement executions that require more memory will be aborted when
they reach the limit. To avoid canceling statements unnecessarily you can also apply a percentage threshold
value which considers the current statement allocation as a proportion of the global memory currently
available. Using this parameter, statements which have exceeded the hard-coded limit may still be executed if
the memory allocated for the statement is within the percentage threshold.

You can view the (peak) memory consumption of a statement in

M_EXPENSIVE_STATEMENTS.MEMORY_SIZE.

Note that M_EXPENSIVE_STATEMENTS.REUSED_MEMORY_SIZE is not used as of SPS 09.

For these options enable_tracking and memory_tracking must be first be enabled in the global.ini file.

You can also create exceptions to these limits for individual users (for example, to ensure an administrator is
not prevented from doing a backup) by setting a different statement memory limit for each individual. For this
option, additionally, resource_tracking must be enabled in the global.ini file.

Procedure

1. Enable statement memory tracking.

In the global.ini file, expand the resource_tracking section and set the following parameters to on:

SAP HANA Troubleshooting and Performance Analysis Guide

32 PUBLIC Root Causes And Solutions
 ○ enable_tracking = on
○ memory_tracking = on
2. In the global.ini file, expand the memorymanager section and set the parameter
statement_memory_limit. Set a statement memory limit in GB (integer values only) with a value
between 1 and some fraction of the global allocation limit.

Note
Values that are too small can block the system from performing critical tasks.

When the statement memory limit is reached, a dump file is created with 'compositelimit_oom' in the
name. The statement is aborted, but otherwise the system is not affected. By default only one dump file is
written every 24 hours. If a second limit hits in that interval, no dump file is written. The interval can be
configured in the memorymanager section of the global.ini file using the oom_dump_time_delta
parameter, which sets the minimum time difference (in seconds) between two dumps of the same kind
(and the same process).
Statements that exceed the limit you have set on a host are stopped by running out of memory.
3. In the global.ini file, expand the memorymanager section and set the parameter
statement_memory_limit_threshold as a percentage of the global allocation limit
(global_allocation_limit).

This parameter provides a means of controlling when the statement_memory_limit is applied. If this
parameter is set, when a statement is issued the system will determine if the amount of memory it
consumes exceeds the defined percentage value of the the overall global_allocation_limit
parameter setting.

This is a way of determining if a particular statement consumes an inordinate amount of memory

compared to the overall system memory available. If so, to preserve memory for other tasks, the
statement memory limit is applied and the statement fails with an exception.
4. To set a user-specific statement limit and exclude a user from the global limit use the ALTER USER
statement as shown here:

ALTER USER <user_name> SET PARAMETER STATEMENT MEMORY LIMIT = <gb>

○ If both a global and a user statement memory limit are set, the user-specific limit takes precedence,
regardless of whether it is higher or lower than the global statement memory limit.
○ If the user-specific statement memory limit is removed the global limit takes effect for the user.
○ The value of the parameter is shown in USER_PARAMETERS (like all other user parameters)

Note
Setting the statement memory limit to 0 will disable any statement memory limit for the user, or, to
reset a user-specific limit use the CLEAR option:

ALTER USER <user_name> CLEAR PARAMETER STATEMENT MEMORY LIMIT

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 33
Results

The following example and scenarios show the effect of applying these settings:

Table 1: Example showing statement memory parameters

Parameter Value

Physical memory 128 GB

global_allocation_limit Default: 90% of the first 64 GB of available physical memory

on the host plus 97% of each further GB; or, in the case of
small physical memory, physical memory minus 1 GB.

statement_memory_limit 1 GB

statement_memory_limit_threshold 60%

Scenario 1:

A statement allocates 2GB of memory and the current used memory size in SAP HANA is 50GB.

● 0,9 * 128GB = 115,2 (global allocation limit)

● 0,6 * 115,2 = 69,12 (threshold in GB)
● 50 GB < 69,12 GB (threshold not reached)

The statement is executed, even though it exceeds the 1GB statement_memory_limit

Scenario 2:

A statement allocates 2GB and the current used memory size in SAP HANA is 70GB

● 70 GB > 69,12 GB (threshold is exceeded)

The statement is cancelled, as the threshold is exceeded, the statement_memory_limit is applied.

Related Information

Parameter Reference: Memory Consumption [page 34]

3.1.5.13.1 Parameter Reference: Memory Consumption

The memorymanager section of the global.ini file contains parameters that allow you to control the
memory consumption of SAP HANA.

You can change the default settings in the configuration editor of the SAP HANA studio (recommended) or
directly in the global.ini system properties file.

These parameters require tracking to be enabled in global.ini [resource_tracking]. Resource tracking creates a
runtime overhead in SAP HANA:

SAP HANA Troubleshooting and Performance Analysis Guide

34 PUBLIC Root Causes And Solutions
● enable_tracking = on
● memory_tracking = on

Note
In a system that supports multitenant database containers, you can configure the global.ini at both the
system level and the database level. Parameters configured at the system level apply to the complete
system and all databases. Parameters configured at the database level apply to the specified database only.

global_allocation_limit - limits the amount of memory that can be used by the system as a whole.

● The parameter global_allocation_limit defines the maximum memory allocation limit in GB.
● The global allocation limit is calculated by default as follows: 90% of the first 64 GB of available physical
memory on the host plus 97% of each further GB. Or, in the case of small physical memory, physical
memory minus 1 GB. A missing entry or a value of 0 results in the system using the default settings.
● Does not require a restart. Available since SPS 08.

Note
In a system that supports multitenant database containers, the global allocation limit configured at the
system layer of the global.ini file is always effective regardless of any value configured at the database
layer.

statement_memory_limit - defines the maximum memory allocation per statement in GB.

● When the statement memory limit is reached, a dump file is created with "compositelimit_oom" in the
name. The statement is aborted, but otherwise the system is not affected.
● The default value is 0 (no limit). Set this parameter to a value between 1 GB and the value of the global
allocation limit.
● Does not require a restart (applies to new statements). Available since SPS 09.

statement_memory_limit_threshold - defines the maximum memory allocation per statement as a

percentage of the global allocation limit.

● If a value for this parameter has been set then the statement memory limit is only applied if the current
SAP HANA memory consumption exceeds the statement memory limit threshold as a percentage of the
global allocation limit.
● The default value is 0% (the statement_memory_limit is always respected). Set this parameter to a value
between 1 GB and the value of the global allocation limit.
● Does not require a restart (applies to new statements). Available since SPS 09.

3.2 CPU Related Root Causes and Solutions

This section covers the troubleshooting of high CPU consumption on the system.

A constantly high CPU consumption will lead to a considerably slower system as no more requests can be
processed. From an end user perspective, the application behaves slowly, is unresponsive or can even seem to
hang.

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 35
Note that a proper CPU utilization is actually desired behavior for SAP HANA, so this should be nothing to
worry about unless the CPU becomes the bottleneck. SAP HANA is optimized to consume all memory and
CPU available. More concretely, the software will parallelize queries as much as possible in order to provide
optimal performance. So if the CPU usage is near 100% for a query execution it does not always mean there is
an issue. It also does not automatically indicate a performance issue.

3.2.1 Indicators of CPU Related Issues

CPU related issues are indicated by alerts issued or in views in the SAP HANA Studio.

The following alerts may indicate CPU resource problems:

● Host CPU Usage (Alert 5)

● Most recent savepoint operation (Alert 28)
● Savepoint duration (Alert 54)

You notice very high CPU consumption on your SAP HANA database from one of the following:

● Alert 5 (Host CPU usage) is raised for current or past CPU usage
● The displayed CPU usage on the overview screen

● The Load graph is currently showing high CPU consumption or shows high consumption in the past

SAP HANA Troubleshooting and Performance Analysis Guide

36 PUBLIC Root Causes And Solutions
3.2.2 Analysis of CPU Related Issues

The following section describes how to analyze high CPU consumption using tools in the SAP HANA studio
tools.

When analyzing high CPU consumption, you need to distinguish between the CPU resources consumed by
HANA itself and by other, non-SAP HANA processes on the host. While the CPU consumption of SAP HANA
will be addressed here in detail, the CPU consumption of other processes running on the same host is not
covered. Such situations are often caused by additional programs running concurrently on the SAP HANA
appliance such as anti-virus and backup software. For more information see SAP Note 1730928.

A good starting point for the analysis is the Overview tab in the SAP HANA studio. It contains a section that
displays SAP HANA CPU usage versus total CPU usage, which includes all processes on the host, and keeps
track of the maximum CPU usage that occurred since the last restart of SAP HANA. If SAP HANA CPU usage is
low while total CPU usage is high, the issue is most likely related to a non-SAP HANA process.

To find out what is happening in more detail, open Performance Threads tab (see Thread Monitoring). In
order to prepare it for CPU time analysis, perform the following steps:

● To switch on resource tracking open the Configuration tab and in the resource_tracking section of the
global.ini file set the following parameters to on.
○ cpu_time_measurement_mode
○ enable_tracking
● Display the CPU Time column by using the Configure Viewer button on the outer right side of the Threads
tab.

The Thread Monitor shows the CPU time of each thread running in SAP HANA in microseconds.. A high CPU
time of related threads is an indicator that an operation is causing the increased CPU consumption.

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 37
 Figure 4: Thread Monitor Showing CPU Time

In order to identify expensive statements causing high resource consumption, turn on the Expensive
Statement trace and specify a reasonable runtime (see Expensive Statements Trace). If possible, add further
restrictive criteria such as database user or application user to narrow down the amount of information traced.
Note that the CPU time for each statement is shown in the column CPU_TIME if resource_tracking is activated.

Another tool to analyze high CPU consumption is the Kernel Profiler. More information about this tool can be
found in Kernel Profiler. Note that setting a maximum duration or memory limit for profiling is good practice
and should be used if appropriate values can be estimated.

To capture the current state of the system for later analysis you can use Full System Info Dump. However,
taking a Full System Info Dump requires resources itself and may therefore worsen the situation. To get a Full
System Info Dump, open Diagnosis Files Diagnosis Information and choose Collect (SQL Procedure) if
the system is up and accepting SQL commands or Collect (Python Script) if it is not.

Related Information

SAP Note 1730928

Thread Monitoring [page 144]
Expensive Statements Trace [page 154]
Kernel Profiler [page 182]

3.2.3 Resolving CPU Related Issues

The first priority in resolving CPU related issues is to return the system to a normal operating state, which may
complicate identifying the root cause

Issue resolution should aim to bring the system back to a sane state by stopping the operation that causes the
high CPU consumption. However, after resolving the situation it might not be possible to find out the actual
root cause. Therefore please consider recording the state of the system under high load for later analysis by
collecting a Full System Info Dump (see Analysis of CPU Related Issues).

Actually stopping the operation causing the high CPU consumption can be done via the Thread Monitor (see
Thread Monitoring). With the columns Client Host, Client IP, Client PID and Application User it is possible to
identify the user that triggered the operation. In order to resolve the situation contact him and clarify the
actions he is currently performing:

SAP HANA Troubleshooting and Performance Analysis Guide

38 PUBLIC Root Causes And Solutions
 Figure 5: Identify Application User

As soon as this is clarified and you agree on resolving the situation, two options are available:

● On the client side, end the process calling the affected threads
● Cancel the operation that is related to the affected threads. To do so, right-click on the thread in the
Threads tab and choose Cancel Operations.

For further analysis on the root cause, please open a ticket to SAP HANA Development Support and attach the
Full System Info Dump, if available.

Related Information

Analysis of CPU Related Issues [page 37]

Thread Monitoring [page 144]

3.2.4 Retrospective Analysis of CPU Related Issues

There are a number of options available to analyze what the root cause of an issue was after it has been
resolved.

A retrospective analysis of high CPU consumption should start by checking the Load graph and the Alerts tab.
Using the alert time or the Load graph, determine the time frame of the high CPU consumption. If you are not
able to determine the time frame because the issue happened too long ago, check the following statistics
server table which includes historical host resource information up to 30 days:

HOST_RESOURCE_UTILIZATION_STATISTICS (_SYS_STATISTICS schema)

With this information, search through the trace files of the responsible process. Be careful to choose the
correct host when SAP HANA runs on a scale-out landscape. The information contained in the trace files will
give indications on the threads or queries that were running during the affected time frame.

If the phenomenon is recurrent due to a scheduled batch jobs or data loading processes, turn on the Expensive
Statements trace during that time to record all involved statements (see Expensive Statements Trace ).
Furthermore, check for concurrently running background jobs like backups and Delta Merge that may cause a
resource shortage when run in parallel. Historical information about such background jobs can be obtained
from the system views:

● M_BACKUP_CATALOG
● M_DELTA_MERGE_STATISTICS

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 39
● A longer history can be found in the statistics server table HOST_DELTA_MERGE_STATISTICS
(_SYS_STATISTICS schema).

Related Information

Expensive Statements Trace [page 154]

M_BACKUP_CATALOG
M_DELTA_MERGE_STATISTICS
HOST_DELTA_MERGE_STATISTICS

3.2.5 Controlling Parallel Execution of SQL Statements

You can apply ini file settings to control the two thread pools SQLExecutors and JobExecutors that control the
parallelism of statement execution.

Caution
The settings described here should only be modified when other tuning techniques like remodeling,
repartitioning, and query tuning have been applied. Modifying the parallelism settings requires a thorough
understanding of the actual workload since they have impact on the overall system behavior. Modify the
settings iteratively by testing each adjustment. For more information, see Understand your Workload.

On systems with highly concurrent workload, too much parallelism of single statements may lead to sub-
optimal performance. Note also that partitioning tables influences the degree of parallelism for statement
execution; in general, adding partitions tends to increase parallelism. You can use the parameters described in
this section to adjust the CPU utilization in the system.

Two thread pools control the parallelism of the statement execution:

● SqlExecutor
This thread pool handles incoming client requests and executes simple statements. For each statement
execution, an SqlExecutor thread from a thread pool processes the statement. For simple OLTP-like
statements against column store as well as for most statements against row store, this will be the only
type of thread involved. With OLTP we mean short running statements that consume relatively little
resources, however, even OLTP-systems like SAP Business Suite may generate complex statements.
● JobExecutor
The JobExecutor is a job dispatching subsystem. Almost all remaining parallel tasks are dispatched to the
JobExecutor and its associated JobWorker threads.

For both SqlExecutor and JobExecutor, a separate limit can be set for the maximum number of threads. For
example, on a system where OLAP workload would normally consume too many CPU resources you can apply
a maximum value to reserve resources for OLTP workload.

Caution
Lowering the value of these parameters can have a drastic effect on the parallel processing of the servers
and reduce the performance of the overall system. Adapt with caution by iteratively making modifications

SAP HANA Troubleshooting and Performance Analysis Guide

40 PUBLIC Root Causes And Solutions
 and testing. For more information, see Understand your Workload and the following subsection, Parameter
Reference: CPU.

Parameters for SQL Executor

The following SqlExecutor parameters are in the sql section of the indexserver.ini.

● sql_executors - The target number of threads that can be used.

● max_sql_executors - The maximum number of threads that can be used. Not set by default so that
new threads are created to handle incoming requests.

Parameters for JobExecutor

The following JobExecutor parameters are in the execution section of the global.ini or
indexserver.ini.

● max_concurrency - The target number of threads that can be used. Note that from SPS12, this
parameter should only need to be modified for multitenant database container installations.
● max_concurrency_hint - Limit concurrency hint even if more active job workers would be available.
● default_statement_concurrency_limit - Used to restrict the actual degree of parallel execution per
connection within a statement.

JobExecutor settings do not solely affect OLAP workload, but also other SAP HANA subsystems (for example,
memory garbage collection, savepoint writes). The JobExecutor executes also operations like table updates
and backups, which were delegated by the SqlExecutor. JobExecutor settings are soft limits, meaning the
JobExecutor can “loan” threads (this applies to the SQL Executor as well), if available, and then fall back to the
maximum number of threads when done.

Tip
In a system that supports multitenant database containers, a reasonable default value for the
max_concurrency parameter is the number of cores divided by the number of tenant databases. Do not
specify a value of 0.

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 41
3.2.6 Controlling CPU Consumption

If the physical hardware on a host is shared between several processes you can use CPU affinity settings to
assign a set of logical cores to a specific SAP HANA process. These settings are coarse-grained and apply on
the OS and process-level.

Prerequisites

Using this workload management option we firstly analyze how the system CPUs are configured and then,
based on the information returned, apply affinity settings in daemon.ini to bind specific processes to logical
CPU cores. Processes must be restarted before the changes become effective. This approach applies
primarily to the use cases of SAP HANA multitenant database containers and multiple SAP HANA instances on
one server; you can use this, for example, to partition the CPU resources of the system by tenant database.

Tip
As an alternative to applying CPU affinity settings you can achieve similar performance gains by changing
the parameter [execution] max_concurrency in the indexserver.ini configuration file. This may be
more convenient and does not require the system to be offline. For more information, see Managing
Resources in Multiple-Container Systems.

To make the changes described here you require access to the operating system of the SAP HANA instance to
run the Linux lscpu command and you require the privilege INIFILE ADMIN.

Context

For Xen and VmWare, the users in the VM guest system see what is configured in the VM host. So the quality
of the reported information depends on the configuration of the VM guest. Therefore SAP cannot give any
performance guarantees in this case.

Procedure

1. Firstly, to confirm the physical and logical details of your CPU architecture, analyze the system using the
lscpu command. This command returns a listing of details of the system architecture. The table which
follows gives a commentary on the most useful values based on an example system with 2 physical chips
(sockets) each containing 8 physical cores. These are hyperthreaded to give a total of 32 logical cores.

# Feature Example Value

1 Architecture x86_64

SAP HANA Troubleshooting and Performance Analysis Guide

42 PUBLIC Root Causes And Solutions
 # Feature Example Value

2 CPU op-mode(s) 32-bit, 64-bit

3 Byte Order LittleEndian

4 CPUs 32

5 On-line CPU(s) list 0-31

6 Thread(s) per core 2

7 Core(s) per socket 8

8 Socket(s) 2

9 NUMA node(s) 2

21 NUMA node0 CPU(s) 0-7,16-23

22 NUMA node1 CPU(s) 8-15,24-31

○ 4-5: This example server has 32 logical cores numbered 0 - 31

○ 6-8: Logical cores ("threads") are assigned to physical cores. Where multiple threads are assigned to a
single physical core this is referred to as 'hyperthreading'. In this example, there are 2 sockets, each
socket contains 8 physical cores (total 16). Two logical cores are assigned to each physical core, thus,
each core exposes two execution contexts for the independent and concurrent execution of two
threads.
○ 9: In this example there are 2 NUMA (Non-uniform memory access) nodes, one for each socket. Other
systems may have multiple NUMA nodes per socket.
○ 21-22: The 32 logical cores are numbered and specifically assigned to one of the two NUMA nodes.

Note
Even on a system with 32 logical cores and two sockets the assignment of logical cores to physical
CPUs and sockets can be different. It is important to collect the assignment in advance before making
changes. A more detailed analysis is possible using the system commands described in the next step.
These provide detailed information for each core including how CPU cores are grouped as siblings.

2. In addition to the lscpu command you can use the set of system commands in the /sys/devices/
system/cpu/ directory tree. For each logical core there is a numbered subdirectory beneath this node (/
cpu12/ in the following examples). The examples show how to retrieve this information and the table
gives details of some of the most useful commands available:

Example
cat /sys/devices/system/cpu/present

cat /sys/devices/system/cpu/cpu12/topology/thread_siblings_list

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 43
 Command Example Output Commentary

present 0-15 The number of logical cores available for scheduling.

cpu12/topology/core_siblings_list 4-7, 12-15 The cores on the same socket.

cpu12/topology/thread_siblings_list 4,12 The logical cores assigned to the same physical core
(hyperthreading).

cpu12/topology/physical_package_id 1 The socket of the current core - in this case cpu12.

3. Based on the results returned you can restrict CPU usage of SAP HANA processes to certain CPUs or
ranges of CPUs. You can do this for the following servers: nameserver, indexserver, compileserver,
preprocessor and xsengine (each server has a section in the daemon.ini file). The examples and
commentary below show the syntax for the ALTER SYSTEM CONFIGURATION commands required. The
changed affinity settings only takes effect after a restart of the affected SAP HANA processes.

Other Linux commands which are relevant here are sched_setaffinity and numactl:
sched_setaffinity limits the set of CPU cores available (by applying a CPU affinity mask) for execution
of a specific process (this could be used, for example, to isolate tenants in a MDC) and numactl controls
NUMA policy for processes or shared memory.

Example
To restrict the nameserver to two logical cores of the first CPU of socket 0 (see line 21 in the example
above), use the following affinity setting:

ALTER SYSTEM ALTER CONFIGURATION ('daemon.ini', 'SYSTEM') SET

('nameserver', 'affinity') = '0,16'

Example
To restrict the preprocessor and the compileserver to all remaining cores (that is, all except 0 and 16)
on socket 0 (see line 21 in the example above), use the following affinity setting:

ALTER SYSTEM ALTER CONFIGURATION ('daemon.ini', 'SYSTEM') SET

('preprocessor', 'affinity') = '1-7,17-23'
ALTER SYSTEM ALTER CONFIGURATION ('daemon.ini', 'SYSTEM') SET
('compileserver', 'affinity') = '1-7,17-23'

Example
To restrict the indexserver to all cores on socket 1 (see line 22 in the example above), use the following
affinity setting:

ALTER SYSTEM ALTER CONFIGURATION ('daemon.ini', 'SYSTEM') SET

('indexserver', 'affinity') = '8-15,24-31'

SAP HANA Troubleshooting and Performance Analysis Guide

44 PUBLIC Root Causes And Solutions
 Example
To set the affinity for two tenant databases called DB1 and DB2 respectively in a multitenant database
container setup, use the following affinity settings::

ALTER SYSTEM ALTER CONFIGURATION ('daemon.ini', 'SYSTEM') SET

('indexserver.DB1', 'affinity') = '1-7,17-23';
ALTER SYSTEM ALTER CONFIGURATION ('daemon.ini', 'SYSTEM') SET
('indexserver.DB2', 'affinity') = '9-15,25-31';

3.3 Disk Related Root Causes and Solutions

This section discusses issues related to hard disks and lack of free space.

Low Disk Space

This issue is usually reported by alert 2 which is issued whenever one of the disk volumes used for data, log,
backup or trace files reaches a critical size.

Use the following tools in the SAP HANA studio to examine the situation and try to free some disk space:

● The Volumes tab

● Open Performance Load .Check Host Disk Used . (See also Load Monitoring )
● Under the System Information tab, open Size of Tables on Disk

More information about the tools can be found in I/O Related Root Causes and Solutions and in the SAP HANA
Administration Guide.

Related Information

Load Monitoring [page 150]

I/O Related Root Causes and Solutions [page 47]
SAP HANA Administration Guide

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 45
3.3.1 Check Internal Disk Full Event (Alert 30)

Alert 30 is issued when it is not possible to write to one of the disk volumes used for data, log, backup or trace
files.

Context

Note that besides running out of disk space, there are more possible causes that may prevent SAP HANA from
writing to disk. All of them will lead to this alert. Example causes include:

● File system quota is exceeded

● File system runs out of inodes
● File system errors (bugs)

Besides doing an analysis via the tools described in Disk Related Root Cause and Solutions, the following
information is helpful too. The commands have to be executed from the command line on the SAP HANA
server:

Procedure

1. Determine the file system type:

df -T
2. Check for disk space using file system specific commands

Option Description

XFS/NFS df
GPFS mmfscheckquota

3. Check if the system is running out of inodes (NFS):

df -i
4. Check quota

Option Description

XFS/NFS quota -v
GPFS mmfscheckquota

Next Steps

If it is not possible to track down the root cause of the alert, contact SAP Support.

SAP HANA Troubleshooting and Performance Analysis Guide

46 PUBLIC Root Causes And Solutions
Related Information

Disk Related Root Causes and Solutions [page 45]

3.4 I/O Related Root Causes and Solutions

This section covers troubleshooting of I/O performance problems. Although SAP HANA is an in-memory
database, I/O still plays a critical role for the performance of the system.

From an end user perspective, an application or the system as a whole runs sluggishly, is unresponsive or can
even seem to hang if there are issues with I/O performance. In the Volumes tab in SAP HANA studio you can
see the attached volumes and which services use which volumes.

Attached volumes In the lower part of the screen you can see details of the volumes, such as files and I/O
statistics.

In certain scenarios data is read from or written to disk, for example during the transaction commit. Most of
the time this is done asynchronously but at certain points in time synchronous I/O is done. Even during
asynchronous I/O it may be that important data structures are locked.

Examples are included in table.

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 47
Table 2:
Scenario Description

Savepoint A savepoint ensures that all changed persistent data since

the last savepoint gets written to disk. The SAP HANA data­
base triggers savepoints in 5 minutes intervals by default.
Data is automatically saved from memory to the data vol­
ume located on disk. Depending on the type of data the
block sizes vary between 4 KB and 16 MB. Savepoints run
asynchronously to SAP HANA update operations. Database
update transactions only wait at the critical phase of the sa­
vepoint, which is usually taking some microseconds.

Snapshot The SAP HANA database snapshots are used by certain op­
erations like backup and system copy. They are created by
triggering a system wide consistent savepoint. The system
keeps the blocks belonging to the snapshot at least until the
drop of the snapshot. Detailed information about snapshots
can be found in the SAP HANA Administration Guide.

Delta Merge The delta merge itself takes place in memory. Updates on
Column Store tables are stored in the delta storage. During
the delta merge these changes are applied to the main stor­
age, where they are stored read optimized and compressed.
Right after the delta merge, the new main storage is per­
sisted in the data volume, that is, written to disk. The delta
merge does not block parallel read and update transactions.

Write Transactions All changes to persistent data are captured in the redo log.
SAP HANA asynchronously writes the redo log with I/O or­
ders of 4 KB to 1 MB size into log segments. Transactions
writing a commit into the redo log wait until the buffer con­
taining the commit has been written to the log volume.

Database restart At database startup the services load their persistence in­
cluding catalog and row store tables into memory, that is,
the persistence is read from the storage. Additionally the
redo log entries written after the last savepoint have to be
read from the log volume and replayed in the data area in
memory. When this is finished the database is accessible.
The bigger the row store is, the longer it takes until the sys­
tem is available for operations again.

Failover (Host Auto-Failover) On the standby host the services are running in idle mode.
Upon failover, the data and log volumes of the failed host
are automatically assigned to the standby host, which then
has read and write access to the files of the failed active
host. Row as well as column store tables (the latter on de­
mand) must be loaded into memory. The log entries have to
be replayed.

SAP HANA Troubleshooting and Performance Analysis Guide

48 PUBLIC Root Causes And Solutions
 Scenario Description

Takeover (System Replication) The secondary system is already running, that is the serv­
ices are active but cannot accept SQL and thus are not usa­
ble by the application. Just like in the database restart (see
above) the row store tables need to be loaded into memory
from persistent storage. If table preload is used, then most
of the column store tables are already in memory. During
takeover the replicated redo logs that were shipped since
the last data transport from primary to secondary have to
be replayed.

Data Backup For a data backup the current payload of the data volumes
is read and copied to the backup storage. For writing a data
backup it is essential that on the I/O connection there are
no collisions with other transactional operations running
against the database.

Log Backup Log backups store the content of a closed log segment.
They are automatically and asynchronously created by
reading the payload from the log segments and writing them
to the backup area.

Database Recovery The restore of a data backup reads the backup content from
the backup device and writes it to the SAP HANA data vol­
umes. The I/O write orders of the data recovery have a size
of 64 MB. Also the redo log can be replayed during a data­
base recovery, that is the log backups are read from the
backup device and the log entries get replayed.

In the below table the I/O operations are listed which are executed by the above mentioned scenarios,
including the block sizes that are read or written.

Table 3:
I/O pattern Data Volume Log Volume (redo log) Backup Medium

Savepoint, WRITE

Snapshot, 4 KB – 16 MB asynchronous
bulk writes, up to 64 MB
Delta merge
(clustered Row Store super
blocks)

Write transactions WRITE

OLTP – mostly 4 KB log write

I/O performance is relevant

OLAP – writes with larger

I/O order sizes

Table load: READ READ

DB Restart, 4 KB – 16 MB blocks, up to
64 MB (clustered Row Store
Failover,
super blocks)
Takeover

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 49
I/O pattern Data Volume Log Volume (redo log) Backup Medium

Data Backup READ WRITE

4 KB – 16 MB blocks, up to in up to 64 MB blocks from

64 MB (clustered Row Store “[data] backup buffer”
super blocks) are asynchro­
nously copied to “[data]
backup buffer” of 512 MB

Log Backup READ WRITE

asynchronously copied to in up to 64 MB blocks from

“[data] backup buffer” of 128 “[data] backup buffer”
MB

Database Recovery WRITE READ READ

4 KB – 16 MB blocks, up to Read block sizes from Read block sizes from

64 MB (clustered Row Store backup file headers à copy backup file headers à copy
super blocks) blocks into “[data] backup blocks into “[data] backup
buffer” of size 512 MB buffer” of size 128 MB

Related Information

SAP HANA Administration Guide

3.4.1 Analyzing I/O Throughput and Latency

When analyzing I/O the focus is on throughput and latency. Monitoring views and SQL statements help with
your analysis.

You can analyze the I/O throughput with this SQL statement:

select v.host, v.port, v.service_name, s.type,

round(s.total_read_size / 1024 / 1024, 3) as "Reads in MB",
round(s.total_read_size / case s.total_read_time when 0 then -1 else
s.total_read_time end, 3) as "Read Througput in MB",
round(s.total_read_time / 1000 / 1000, 3) as "Read Time in Sec",
trigger_read_ratio as "Read Ratio",
round(s.total_write_size / 1024 / 1024, 3) as "Writes in MB",
round(s.total_write_size / case s.total_write_time when 0 then -1 else
s.total_write_time end, 3) as "Write Througput in MB",
round(s.total_write_time / 1000 / 1000, 3) as "Write Time in Sec" ,
trigger_write_ratio as "Write Ratio"
from "PUBLIC"."M_VOLUME_IO_TOTAL_STATISTICS_RESET" s, PUBLIC.M_VOLUMES v
where s.volume_id = v.volume_id
and type not in ( 'TRACE' )
and v.volume_id in (select volume_id from m_volumes where service_name =
'indexserver')
order by type, service_name, s.volume_id;

SAP HANA Troubleshooting and Performance Analysis Guide

50 PUBLIC Root Causes And Solutions
The system view M_VOLUME_IO_TOTAL_STATISTICS_RESET is used to get the size of reads and writes and
the throughput in MB for the indexserver since the last reset of the counters.

The Ratio fields indicate bad performance, if they are drifting towards 1. They should tend towards 0.

Explanation of Ratio: I/O calls are executed asynchronously; that is the thread does not wait for the order to
return. A ratio close to 0 says that the thread does not wait at all; a ratio close to 1 means that the thread waits
until I/O request is completed because the asynchronous call is blocked (time for triggering I/O time for I/O
completion).

More information can be found in SAP Note 1930979.

It is possible to reset the view and analyze the I/O throughput for a certain time frame by running the reset
command below and query again after the desired time frame.

alter system reset monitoring view M_VOLUME_IO_TOTAL_STATISTICS_RESET;

The latency is important for LOG devices. To analyze the latency, use:

select host, port type,

round(max_io_buffer_size / 1024, 3) "Maximum buffer size in KB",
trigger_async_write_count,
avg_trigger_async_write_time as "Avg Trigger Async Write Time in
Microsecond",
max_trigger_async_write_time as "Max Trigger Async Write Time in
Microsecond",
write_count, avg_write_time as "Avg Write Time in Microsecond",
max_write_time as "Max Write Time in Microsecond"
from "PUBLIC"."M_VOLUME_IO_DETAILED_STATISTICS_RESET"
where type = 'LOG'
and volume_id in (select volume_id from m_volumes where service_name =
'indexserver')
and (write_count <> 0 or avg_trigger_async_write_time <> 0);

With this statement you get the log write wait time (for data of type LOG) with various buffer sizes written by
the indexserver. All measures are the periods of time between enqueueing and finishing a request.

Related Information

SAP Note 1930979

M_VOLUME_IO_TOTAL_STATISTICS_RESET

3.4.2 Savepoint Performance

To perform a savepoint write operation, SAP HANA needs to take a global database lock. This period is called
the “critical phase” of a savepoint. While SAP HANA was designed to keep this time period as short as
possible, poor I/O performance can extend it to a length that causes a considerable performance impact.

Savepoints are used to implement backup and disaster recovery in SAP HANA. If the state of SAP HANA has
to be recovered, the database log from the last savepoint will be replayed.

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 51
You can analyze the savepoint performance with this SQL statement:

select start_time, volume_id,

round(duration / 1000000) as "Duration in Seconds",
round(critical_phase_duration / 1000000) as "Critical Phase Duration in
Seconds",
round(total_size / 1024 / 1024) as "Size in MB",
round(total_size / duration) as "Appro. MB/sec",
round (flushed_rowstore_size / 1024 / 1024) as "Row Store Part MB"
from m_savepoints
where volume_id in ( select volume_id from m_volumes where service_name =
'indexserver') ;

This statement shows how long the last and the current savepoint writes took/are taking. Especially the
critical phase duration, in which savepoints need to take a global database lock, must be observed carefully.

The critical phase duration should not be longer than a second. In the example below the times are
significantly higher due to I/O problems.

Figure 6: Savepoints

The following SQL shows a histogram on the critical phase duration:

select
to_char(SERVER_TIMESTAMP,'yyyy.mm.dd') as "time",
sum(case when (critical_phase_duration <= 1000000) then 1 else 0
end) as "<= 1 s",
sum(case when (critical_phase_duration > 1000000 and critical_phase_duration
<=2000000) then 1 else 0
end) as "<= 2 s",
sum(case when (critical_phase_duration > 2000000 and critical_phase_duration
<=3000000) then 1 else 0
end) as "<= 3 s",
sum(case when (critical_phase_duration > 3000000 and critical_phase_duration
<=4000000) then 1 else 0
end) as "<= 4 s",
sum(case when (critical_phase_duration > 4000000 and critical_phase_duration
<=5000000) then 1 else 0
end) as "<= 5 s",
sum(case when (critical_phase_duration > 5000000 and critical_phase_duration
<=10000000) then 1 else 0
end) as "<= 10 s",
sum(case when (critical_phase_duration > 10000000 and critical_phase_duration
<=20000000) then 1 else 0
end) as "<= 20 s",
sum(case when (critical_phase_duration > 20000000 and critical_phase_duration
<=40000000) then 1 else 0
end) as "<= 40 s",
sum(case when (critical_phase_duration > 40000000 and critical_phase_duration
<=60000000) then 1 else 0
end) as "<= 60 s",
sum(case when (critical_phase_duration > 60000000 ) then 1 else 0
end) as "> 60 s",

SAP HANA Troubleshooting and Performance Analysis Guide

52 PUBLIC Root Causes And Solutions
 count(critical_phase_duration) as "ALL"
from "_SYS_STATISTICS"."HOST_SAVEPOINTS"
where volume_id in (select volume_id from m_volumes where service_name =
'indexserver')
and weekday (server_timestamp) not in (5, 6)
group by to_char(SERVER_TIMESTAMP,'yyyy.mm.dd')
order by to_char(SERVER_TIMESTAMP,'yyyy.mm.dd') desc;

Figure 7: Savepoint Histogram

The performance of the backup can be analyzed with this statement:

select mbc.backup_id,
SECONDS_BETWEEN (mbc.sys_start_time, mbc.sys_end_time) seconds,
round(sum(backup_size) / 1024 / 1024 / 1024,2) size_gb,
round(sum(backup_size) / SECONDS_BETWEEN (mbc.sys_start_time, mbc.sys_end_time) /
1024 / 1024, 2) speed_mbs
from m_backup_catalog_files mbcf , m_backup_catalog mbc
where mbc.entry_type_name = 'complete data backup'
and mbc.state_name = 'successful'
and mbcf.backup_id = mbc.backup_id
group by mbc.backup_id, mbc.sys_end_time, mbc.sys_start_time order by
mbc.sys_start_time

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 53
3.5 Configuration Parameter Issues

The SAP HANA database creates alerts if it detects an incorrect setting for any of the most critical
configuration parameters.

The following table lists the monitored parameters and related alerts.

Table 4:
Alert ID Alert Name Parameter Further Information

10 Delta merge (mergedog) Indexserver.ini – mergedog - Delta Merge

configuration active

16 Lock wait timeout configura­ Indexserver.ini – transaction Transactional Problems

tion – lock_wait_timeout

32 Log mode legacy Global.ini – persistence – Issues with Configuration Pa­

log_mode rameter log_mode (Alert 32
and 33)

33 Log mode overwrite Global.ini – persistence – Issues with Configuration Pa­

log_mode rameter log_mode (Alert 32
and 33)

To check for parameters that are not according to the default settings, the following SQL statement can be
used.

select a.file_name, b.layer_name, b.tenant_name, b.host, b.section, b.key,

a.value as defaultvalue, b.currentvalue from sys.m_inifile_contents a
join ( select file_name, layer_name, tenant_name, host, section, key,
value as currentvalue from sys.m_inifile_contents b where layer_name
<> 'DEFAULT' )
b on a.file_name = b.file_name and a.section = b.section and a.key = b.key
and a.value <> b.currentvalue

Note
Default values of parameters may change when updating the SAP HANA database with a new revision.
Custom values on the system level and on the host level will not be affected by such updates.

Correcting Parameter Settings

Usually alerts on incorrect parameter settings include information about correct setting of the parameter. So,
unless you have received a specific recommendation from SAP to change the parameter to another value, you
can fix the issue by changing the parameter from the Configuration tab of SAP HANA studio. You can filter on
the parameter name to find it. In most cases the suggested correct value will be the default value.

Note
Make sure that you change the parameter in the correct ini-file and section, since the parameter name itself
may be not unique.

SAP HANA Troubleshooting and Performance Analysis Guide

54 PUBLIC Root Causes And Solutions
Most of the parameters can be changed online and do not require any further action. Exceptions for common
parameters are documented in SAP Note 1891582.

Related Information

Delta Merge [page 56]

Transactional Problems [page 74]
Issues with Configuration Parameter log_mode (Alert 32 and 33) [page 55]
SAP Note 1891582

3.5.1 Issues with Configuration Parameter log_mode (Alert

32 and 33)

Alerts 32 and 33 are raised whenever the write mode to the database log is not set correctly for use in
production.

Context

To ensure point-in-time recovery of the database the log_mode parameter must be set to ‘normal’ and a data
backup is required.

The following steps are recommended when facing this alert:

Procedure

1. Change the value of the parameter log_mode in SAP HANA studio to normal
2. Schedule an initial data backup
3. Test successful completion of the backup
4. Restart the database
5. Backup the database configuration
For information on how to perform a backup of database configuration files see SAP Note 1651055.
6. Schedule a regular data backup

Related Information

SAP Note 1651055

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 55
3.6 Delta Merge

This section covers troubleshooting of delta merge problems.

The Column Store uses efficient compression algorithms to keep relevant application data in memory. Write
operations on the compressed data are costly as they require reorganizing the storage structure and
recalculating the compression. Therefore write operations in Column Store do not directly modify the
compressed data structure in the so called main storage. Instead, all changes are at first written into a
separate data structure called the delta storage and at a later point in time synchronized with the main
storage. This synchronization operation is called delta merge.

From an end user perspective, performance issues may occur if the amount of data in the delta storage is
large, because read times from delta storage are considerably slower than reads from main storage.

In addition the merge operation on a large data volume may cause bottleneck situations, since the data to be
merged is hold twice in memory during the merge operation.

The following alerts indicate an issue with delta merges:

● Delta merge (mergedog) configuration (Alert 10)

● Size of delta storage of Column Store tables (Alert 29)

3.6.1 Inactive Delta Merge

In case the delta merge is set to inactive, Alert 10 Delta Merge (mergedog) Configuration is raised. In a
production system this alert needs to be handled with very high priority in order to avoid performance issues.

Context

Whenever issues with delta merge are suspected, this alert should be checked first. An inactive delta merge
has a severe performance impact on database operations.

Figure 8: Delta Merge Set To Inactive

Procedure

1. Check the current parameter value in the Configuration tab of SAP HANA studio and filter for the
parameter mergedog.
Check the value of active in the mergedog section of the indexserver.ini.

SAP HANA Troubleshooting and Performance Analysis Guide

56 PUBLIC Root Causes And Solutions
 Figure 9: Check Mergedog Active
2. To correct the value, double-click on active and choose Restore Default.

This will delete all custom values on system and host level and restore the default value system-wide.

Figure 10: Restore Defaults

Note
Depending on the check frequency (default frequency: 15 minutes) the alert will stay in the Alert inbox
until the new value is recognized the next time the check is run.

3.6.2 Retrospective Analysis of Inactive Delta Merge

Retrospective analysis of the root cause of the parameter change that led to the configuration alert requires
the activation of an audit policy in SAP HANA that tracks configuration changes

Other sources of information are external tools (for example, SAP Solution Manager) that create a snapshot of
configuration settings at regular intervals.

For details about configuring security auditing and for analyzing audit logs, refer to the SAP HANA Security
Guide.

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 57
Related Information

SAP HANA Security Guide

3.6.3 Indicator for Large Delta Storage of Column Store

Tables

If the delta storage of a table gets too large, read operations on the table will slow down. This usually results in
degraded performance of queries reading from the affected table.

When the delta storage of a table gets too large, the Alert Size of delta storage of Column Store tables (Alert
29) can be raised.

Alert 29 is raised when the amount of memory consumed by the delta storage exceeds the configured
thresholds. The thresholds can be customized in the SAP HANA studio to take into account the configured size
of the delta storage. Note that if the alerts are not configured properly, the symptoms can occur without
raising an alert, or there may be no symptoms, even though an alert is raised. For each affected table a
separate alert is created.

Usually this problem occurs because of mass write operations (insert, update, delete) on a column table. If the
total count of records (record count * column count) in the delta storage exceeds the threshold of this alert
before the next delta merge, the alert Check delta storage record count * table column count will be triggered.

Corrective action needs to be taken in one of the following areas:

● Change of an application
● Changed partitioning of the table
● Configuration of delta merge

3.6.4 Analyze Large Delta Storage of Column Store Tables

Analyze and interpret issues related to delta storage with help from alerts in SAP HANA studio.

Procedure

1. If an alert was raised, go to the Alerts Tab in the SAP HANA studio and filter for "delta storage".
Check if the alert is raised for a small number of tables or if it is raised for multiple tables. Focus on tables
where the alert has high priority. Alerts raised with low or medium priority usually don’t need immediate
action, but should be taken as one indicator for checking the sizing. Also these alerts should be taken into
account when specific performance issues with end-user operations on these tables are reported, since
read-access on delta storage may be one reason for slow performance.

SAP HANA Troubleshooting and Performance Analysis Guide

58 PUBLIC Root Causes And Solutions
 Figure 11: Check Alert Details
2. Double-click on an alert and check the alert details about its previous occurrences.
a. If the alert occurred several times, check since when this started.
b. Check whether it occurs regularly at a certain time.
This may indicate a specific usage pattern from application side that might have room for
optimization. For example, when many inserts and deletes are performed during a load process, it
might be possible to replace these operations with a suitable filter in the source system. To determine
the usage of the table by applications, the data in the Expensive Statements Trace and Load monitor
can be employed.

3. Check the time stamp of the alert if it is current, then start with checking current attributes of this table.
Information regarding the delta merge operation on specific tables can be obtained from the system view
M_CS_TABLES.

Figure 12: M_CS_TABLES Information

SELECT * FROM SYS.M_CS_TABLES where table_name='<name of table>' and

schema_name='<name of schema>';

If no alert was raised, you can check for the tables with the most records in the delta.

SELECT * FROM SYS.M_CS_TABLES where record_count>0 order by

raw_record_count_in_delta desc;

4. Check the following attributes:

○ LAST_MERGE_TIME
○ MERGE-COUNT
○ READ_COUNT, WRITE_COUNT
○ RECORD_COUNT
○ RAW_RECORD_COUNT_IN_MAIN

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 59
 ○ RAW_RECORD_COUNT_IN_DELTA
○ MEMORY_SIZE_IN_MAIN
○ MEMORY_SIZE_IN_DELTA
a. If MERGE_COUNT is high then this is an indicator that the delta merge works properly, while a low
MERGE_COUNT suggests a need for corrective action.

A large difference between RAW_RECORD_COUNT_IN_MAIN and RECORD_COUNT suggests that the

table has not been compressed properly. Note that compression is not triggered when a merge is
triggered from an SQLScript, but only in case of AUTO-, SMART- or CRITICAL- Merge.

A high WRITE_COUNT suggests that many insert, update and delete operations occur. If the
occurrence of the delta merge problem is rare, then it usually will be sufficient to trigger the merge for
this table manually. See Perform a Manual Delta Merge Operation in the SAP HANA Administration
Guide.
b. If there are many deleted records, it is also required to trigger a compress of the table with the
following command:

UPDATE <name of table> WITH PARAMETERS('OPTIMIZE_COMPRESSION'='YES');

c. Confirm the delta merge operation has succeeded in the following ways:
Open the table definition in the table editor and on the Runtime Information tab and check the relevant
values:
○ LAST_MERGE_TIME
○ MERGE_COUNT
○ RAW_RECORD_COUNT_IN_DELTA
○ LAST_COMPRESSED_RECORD_COUNT
5. If WRITE_COUNT is low, check the threshold value of "Check currently utilized percentage of main
memory" in Configure Check Settings of the Alerts tab in SAP HANA studio.
Unless other recommendation has been provided by SAP the default values shall be applied to the system.
Default values are:
○ Low: 800,000,000
○ Medium: 1,600,000,000
○ High: 4,000,000,000
If you find other (lower) settings, then it is likely that the alert occurred due to incorrect configuration of
the alerting rather than due to issues with tables, applications or delta merge functions. To resolve this,
change the settings back to the default values:

SAP HANA Troubleshooting and Performance Analysis Guide

60 PUBLIC Root Causes And Solutions
 Figure 13: Configure Check Settings
6. If problems with the delta storage re-occur frequently for a specific table, check Merge Statistics for this
table. This can be done in System Information > Merge Statistics, where you can put a filter on the table
name and schema name.

Figure 14: Merge Statistics in SAP HANA studio

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 61
 Alternatively you can run the following SQL statement and perform the following checks:

select * from SYS.M_DELTA_MERGE_STATISTICS where table_name='<name of table>'

and schema_name='<name of schema>';

a. Check column SUCCESS for records with value other than TRUE.
b. Check the column LAST_ERROR for records with value other than 0. A typical error is 2048 and
ERROR_DESCRIPTION shows error 2484 which indicates that there was not enough memory to
compress the table after the merge.
For other error codes please refer to the SAP HANA Administration Guide.
c. Check the columns START_TIME, EXECUTION_TIME, MOTIVATION and MERGED_DELTA_RECORDS.
For cases where MERGED_DELTA_RECORDS becomes excessively large the trigger function for the
MOTIVATION type should be reviewed and the LOAD should be analyzed for that time frame
( Performance Load ). A value of MERGED_DELTA_RECORDS = -1 suggests that no records were
merged but that a compression optimization was performed.
7. If you need to analyze the delta merge statistics for a longer period, than use the equivalent select on table
HOST_DELTA_MERGE_STATISTICS of the statistics server:

SELECT * FROM _SYS_STATISTICS.HOST_DELTA_MERGE_STATISTICS where

table_name='<name of table>' and schema_name='<name of schema>';

The delta merge configuration can checked in the SAP HANA studio by opening Configuration
indexserver.ini mergedog
Figure 15: Merge Dog Parameter

Since the default value for the frequency of delta merges is already 1 minute (check_interval = 60.000
ms), optimization with regards to memory consumption can only be done by adjusting the decision
function of the corresponding merge type and the corresponding priority function. However, changes
should be done very carefully and always with involvement of experts from SAP. Parameters of the
functions are documented in the SAP HANA Administration Guide.

Related Information

SAP HANA Administration Guide

M_CS_TABLES

SAP HANA Troubleshooting and Performance Analysis Guide

62 PUBLIC Root Causes And Solutions
3.6.5 Failed Delta Merge

If many cases are identified where auto merge has failed, the error codes need to be analyzed in more detail.
Note that the merge only failed if SUCCESS is not TRUE. In any other case the error code describes a non-
critical condition during a successful merge.

To analyze the error codes, you should increase the trace level to INFO for the components mergedog and
mergemonitorin the INDEXSERVER section of the Database Trace.

To change the trace configuration go to the Trace Configuration tab in SAP HANA studio and change the
configuration of the Database Trace.

Note
You need to select the Show all components checkbox to display the mentioned trace components.

Figure 16: Database Trace

The following table lists error codes and typical corrective actions.

Table 5: Error Codes

Error Description Recommended Action

1999 General error (no further information Check the indexserver trace for more
available) errors regarding the exception

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 63
Error Description Recommended Action

2450 Error during merge of delta index oc­ Check in diagnosis files for an Out-Of-
curred Memory dump that occurred during
the delta merge operation

2458 Table delta merge aborted as cancel No action required.

was manually requested by a kill ses­
sion call.

2480 The table in question is already being No action required.

merged.

2481 There are already other smart merge No action required.

requests for this table in the queue.

2482 The delta storage is empty or the evalu­ No further action required if this occurs
ation of the smart merge cost function
occasionally.
indicated that a merge is not neces­
sary. If it happens frequently:

Check M_DELTA_MERGE_STATISTICS
and review smart merge cost function
with SAP experts.

(parameter
smart_merge_decision_func)

2483 Smart merge is not active (parameter Change the parameter

smart_merge_enabled=no) smart_merge_enabled=yes)

2484 Memory required to optimize table ex­ No further action required if this occurs
ceeds heap limit (for failed compres­
occasionally.
sion optimization operations,
TYPE=SPARSE, SUCCESS=FALSE). If it happens frequently:

A) Analyze change operations on the

table and consider table partitioning to
minimize the size of the delta storage.
If no knowledge about application is
available, Hash Partitioning with a size
of 500.000.00 records is a good initial
choice.

B) Analyze change operations on the

table and consider adjusting the pa­
rameter
auto_merge_decision_func

C) Increase delta storage

D) Review sizing

6900 Attribute engine failed Internal error. Check the indexserver

trace for more errors regarding the ex­
ception.

SAP HANA Troubleshooting and Performance Analysis Guide

64 PUBLIC Root Causes And Solutions
 Error Description Recommended Action

29020 ltt::exception caught while operating on Internal error. Check the indexserver
$STORAGEOBJECT$ trace for more errors regarding the ex­
ception.

3.6.6 Delta Storage Optimization

Table partitioning allows you to optimize the size of tables in memory and their memory consumption as each
partition has its own delta storage.

The memory consumption of a table in memory during a merge operation depends on the number of records,
the number and memory size of columns and the memory size of the table. While the number of records can
be kept low by triggering a smart merge from the application, optimization with regards to the size of the table
can be achieved by table partitioning. This is due to the fact that each partition holds a separate delta storage.
When a merge is performed, the data from the main storage has to be loaded into memory which is a
considerably less amount when only a single partition is handled rather than the full table.

When considering partitioning it is recommended to analyze the typical usage of this table. Partitions should
be created in a way that avoids as much as possible that single statements need to access multiple partitions.
If no application knowledge is available, then hash partitioning with a partition size of about 500.000.000
records is a good initial choice.

See, Table Partitioning in the SAP HANA Database in the SAP HANA Administration Guide.

Related Information

SAP HANA Administration Guide

3.7 Security-Related Issues

This section looks at issues to do with security like licensing, authorization, and authentication.

3.7.1 System Locked Due to Missing, Expired, or Invalid

License

New installations of SAP HANA are equipped with a temporary license that expires after 90 days. To keep the
system functional after this period, you have to install a permanent license.

Improper licensing may lead to a lockdown of your SAP HANA system. In this case, the only allowed action is
to install a valid license.

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 65
The system goes into lockdown in the following situations:

● Your first temporary license of 90 days has expired.

● Your permanent license has expired and you do not renew it within 28 days.
● An old backup was used for recovery and the license key in the backup has expired in the meantime.
● The installed license key is an enforced license key and current memory consumption exceeds the amount
specified in the license key. Note that such licenses are only used in some scenarios.
● You have deleted all license keys installed in your database.

For more information, see Managing SAP HANA Licenses in the SAP HANA Administration Guide.

Related Information

SAP HANA Administration Guide

3.7.2 License Problem Identification and Analysis

The first signs of problems related to licensing will be visible by Alert 31 or Alert 44 being issued.

To check your current license using SAP HANA studio, right click on a system in the Systems view, choose
Properties and then License.

Alternatively, you can retrieve the same information using SQL:

select * from m_license;

The M_LICENSE system view provides you with the following information:

● License data:
○ SID
○ Hardware key
○ Installation number
○ System number
○ Product limit (licensed amount of memory)
○ Validity start date
○ Expiration date
○ Last successful check date
● License status (permanent, valid, or enforced)

Note that in case of system lockdown, only SID and hardware key are displayed. Information on previously
installed licenses is available.

Note
To be able to query license information, you must have the system privilege LICENSE ADMIN.

SAP HANA Troubleshooting and Performance Analysis Guide

66 PUBLIC Root Causes And Solutions
Related Information

SAP HANA SQL and System Views Reference

3.7.3 Resolution of License Issues

If your license becomes invalid, you need to install a new license.

You can install a new license either in the SAP HANA studio or using SQL.

Note
To install a license key, you need the LICENSE ADMIN system privilege.

You install a license key with the following SQL statement:

SET SYSTEM LICENSE '<license file content goes here, line breaks matter>';

Note
Line breaks are essential for interpretation of the license key text, hence they must not be removed. If you
use the command line tool SAP HANA HDBSQL to install the license, make sure to enable multi-line
statement support (command line option -m or \mu ON when within SAP HANA HDBSQL).

The command will fail if the license key has a different installation number or system number than the current
ones in the database instance. If you have successfully installed a license but your system is still locked down,
check the following:

● The current system time is within the validity period of the license.
● Your installed license key is correct, in particular, the M_LICENSE view displays only one row with a valid
license for the product SAP HANA.
● The SAP Notes in the Related Links section.

For more detailed information about how to install a license key, see Install a Permanent License in the SAP
HANA Administration Guide.

Related Information

SAP Note 1704499 - System Measurement for License Audit

SAP Note 1634687 - License request and installation for SAP HANA database
SAP Note 1699111 - License key update in SAP HANA prior to Rev.26
SAP HANA Administration Guide

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 67
3.7.4 Troubleshooting Authorization Problems

SAP HANA implements its authorization concept based on the entities user, privilege, and role.

General Analysis

The system view EFFECTIVE_PRIVILEGES is useful for checking the privileges of a specific user. It includes
information about all privileges granted to a specific user (both directly and indirectly through roles), as well as
how the privileges were obtained (GRANTOR and GRANTOR_TYPE column).

Figure 17: Output of Effective Privileges

For more information about using this view and other system views related to authorization, see System Views
for Verifying Users' Authorization.

For more information about the authorization concept in SAP HANA, see the SAP HANA Security Guide.

Related Information

SAP HANA Security Guide

SAP HANA Administration Guide
SAP HANA SQL and System Views Reference

SAP HANA Troubleshooting and Performance Analysis Guide

68 PUBLIC Root Causes And Solutions
3.7.4.1 Troubleshoot the Error "Insufficient Privilege: Not
Authorized"

If the error Insufficient privilege: Not authorized occurs during statement execution, you need to
find out which privileges the user is missing and then grant them to the user.

Prerequisites

You have the system privilege TRACE ADMIN.

Procedure

1. On the Trace Configuration tab of the Administration editor, set the database trace level for the
component authorization of the indexserver service to INFO.

Note
The component is not visible by default. To see it, choose Choose All Components.

2. Execute the statement that triggered the error.

3. Set the database trace level for the component authorization of indexserver service back to DEFAULT.
4. On the Diagnosis Files tab of the Administration editor, examine the indexserver trace to find out about the
failed authorization check.
Usually, you will find something like: User <user_name> is missing … privilege for
<object_type> <object_name>. Use this information to grant the missing privilege. If the user cannot
access a view due to a missing analytic privilege, the trace will also list all relevant analytic privileges that
have not been granted to the user.

Related Information

SAP HANA Administration Guide

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 69
3.7.4.2 Troubleshoot the Display of Unrestricted or
Incorrect Results for a View Secured with Analytic
Privileges

If a user has unrestricted access to a view or sees results that he should not, even though he has been granted
an analytic privilege, you need to determine which privileges have been granted to the user and whether or not
they are correct.

Prerequisites

To troubleshoot this issue, you require the following system privileges:

● CATALOG READ
● TRACE ADMIN

Procedure

● Check which analytic privileges have been granted to the user using the system view
EFFECTIVE_PRIVILEGES.
Execute the following SQL statement:

SELECT * FROM EFFECTIVE_PRIVILEGES WHERE USER_NAME = '<user>' AND OBJECT_TYPE

= 'ANALYTICALPRIVILEGE';

In particular, verify that the user does not have the analytic privilege _SYS_BI_CP_ALL. This analytic
privilege potentially allows a user to access all the data in all activated views, regardless of any other
analytic privileges that apply. Usually, the user will have this analytic privilege through a role, for example,
MODELING.

Caution
The MODELING role is very privileged and should not be granted to users, particularly in production
systems. The MODELING role should only be used as a template.

● Identify wrong filters specified in the analytic privileges granted to the user.
Information about filter conditions generated from the relevant analytic privileges can be traced in the
indexserver trace file. This can help you to identify wrong filters specified in the analytic privileges granted
to the user.

On the Trace Configuration tab of the Administration editor, set the database trace level for the
component analyticprivilegehandler of the indexserver service to DEBUG.

SAP HANA Troubleshooting and Performance Analysis Guide

70 PUBLIC Root Causes And Solutions
Related Information

SAP HANA Administration Guide

SAP HANA SQL and System Views Reference
SAP HANA Security Guide

3.7.4.3 Troubleshoot the Error "Insufficient privilege: Not

authorized" Although User Has Analytic Privileges

Even if a user has the correct analytic privileges for a view, he still may receive the error Insufficient
privilege: Not authorized if there is an issue with privileges at another level.

Prerequisites

To troubleshoot this issue, you require the following system privileges:

● CATALOG READ
● TRACE ADMIN

Procedure

● Verify that the _SYS_REPO user has all required privileges (for example, SELECT) with GRANT OPTION on
the base tables of the view.
You can do this by selecting from the EFFECTIVE_PRIVILEGES system view:

SELECT * FROM EFFECTIVE_PRIVILEGES WHERE USER_NAME = '_SYS_REPO';

● Verify that the analytic privileges required for any underlying views have been granted to the user.
If the view is a top-level view (calculation view) with underlying views, the granted analytic privilege grants
access only to this top-level view. Analytic privileges are required for all underlying views. Note that
analytic privileges have to contain at least a view attribute with or without filter condition in order to grant
access to the view.

You can verify a user's privilges by selecting from the EFFECTIVE_PRIVILEGES system view:

SELECT * FROM EFFECTIVE_PRIVILEGES WHERE USER_NAME = '<user>' AND OBJECT_TYPE

= 'ANALYTICALPRIVILEGE';

● If the analytic privilege uses a database procedure to define dynamic value filters at runtime, check for
errors in the execution of the underlying procedure.
To find out the actual error during procedure execution for analytical privileges, check the
indexserver_alert_<host>.trc trace file (accessible on the Diagnosis Files tab of the Administration
editor).

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 71
3.7.4.4 Troubleshoot the Error "Invalidated View" During
SELECT Statement Execution
A user may receive the error Invalidated view when executing a SELECT statement against a view that
was activated from the repository. In addition, thee data preview for an activated view does not show any data.

Prerequisites

To troubleshoot this issue, you require the following system privileges CATALOG READ.

Procedure

● Verify that the _SYS_REPO user has all required privileges (for example, SELECT) on all base objects (for
example, tables) of the view.
You can do this by selecting from the EFFECTIVE_PRIVILEGES system view:

SELECT * FROM EFFECTIVE_PRIVILEGES WHERE USER_NAME = '_SYS_REPO';

3.7.5 Troubleshooting Problems with User Name/Password

Authentication
Common problems with regards to authentication are related to incorrect or expired passwords.

User administrators can change users' password in the User editor of the SAP HANA studio.
Figure 18: User Editor

SAP HANA Troubleshooting and Performance Analysis Guide

72 PUBLIC Root Causes And Solutions
For more information about managing users in the User editor, see Security Administration in the SAP HANA
Administration Guide.

Related Information

SAP HANA Administration Guide

3.7.5.1 Resetting the Password for the SYSTEM User

In case the password for the SYSTEM user is not available anymore, it can be reset.

As a prerequisite, <sid>adm access to the (primary) SAP HANA host is required. Follow the procedure as
described in detail in the SAP HANA Administration Guide.

Related Information

SAP HANA Administration Guide

3.7.5.2 Troubleshoot the Error "User is locked"

A user receives the error User is locked after too many failed log on attempts.

Prerequisites

You have system privilege USER ADMIN.

Context

An example of this error might look like:

Error "user is locked; try again later: lock time is 1440 minutes; user is locked
until 2014-05-28 21:42:24.12214212" (the time is given in UTC).

Most likely, the user logged on too many times with the wrong password. The default maximum number of
failed logon attempts is 6. This is defined by the password policy parameter
maximum_invalid_connect_attempts.

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 73
For more information about this and other password policy parameters, see Password Policy Configuration
Options in the SAP HANA Security Guide.

Procedure

Reset the invalid connect attempts with the following SQL statement:

ALTER USER <user> RESET CONNECT ATTEMPTS;

The user can now log on again.

Related Information

SAP HANA Security Guide

3.8 Transactional Problems

This section covers troubleshooting of transaction problems. From an end user perspective, an application
runs sluggishly, is unresponsive or can even seem to hang if there are issues with uncommitted transactions,
long-lived cursors blocking garbage collection, a high number of active versions or blocked transactions.

3.8.1 Blocked Transactions

Blocked transactions are write transactions that are unable to be further processed because they need to
acquire transactional locks (record or table locks) which are currently held by another write transaction. Note
that transactions can also be blocked waiting for physical resources like network or disk. Those situations are
not covered in this section.

3.8.1.1 Identify and Assess Blocked Transaction Issues

The first signs of blocked transactions are poor application response or alerts 49 or 59 are raised.

The initial indicators of blocked transactions are given by:

● Users reporting bad application responsiveness

● Alert 49 - Long-running blocking situations
● Alert 59 - Percentage of transactions blocked

SAP HANA Troubleshooting and Performance Analysis Guide

74 PUBLIC Root Causes And Solutions
To confirm the database performance is harmed by blocked transactions, you should check the following SAP
HANA studio monitors under the Performance tab:

Load Monitor

Figure 19: SAP HANA Studio Load Monitor

The Blocked Transactions graph shows how many blocked transactions currently exist and existed in the past
to a certain extent. See Load Monitoring.

Job Progress Monitor

To further track down the issue, look at the Job Progress monitor. It shows currently running SAP HANA
background processes like Delta Table Merge. Since the Delta Table Merge needs to lock tables to proceed, it
is a common cause for blocked transactions. Another job display by this monitor is the savepoint write which
needs to pull a global database lock in its critical phase. See Job Progress Monitoring and Savepoint
Performance.

Session Monitor

The Session Monitor lists all currently opened SQL sessions (meaning user connections). In the context of
blocked transaction troubleshooting, the columns “Blocked by Connection Id” and “Blocks No. of
Transactions” are of special interest. The first tells you whether the session is blocked by another session and
identifies the ID of the blocking one. The latter gives you the corresponding information if a session blocks
other sessions, and how many transactions are affected. See Session Monitoring.

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 75
 Figure 20: SAP HANA Session Monitor

Blocked Transaction Monitor

The Blocked Transaction Monitor is the next drill down step. It only lists those transactions that are currently
blocked. The ordering is done via a blocking/blocked relation. That means transactions that are blockers are
highlighted. Directly beneath the blocked transaction are displayed:

Figure 21: Blocked Transaction Monitor

Example: In the figure above, you see transaction 126 (green) blocking multiple other transactions (red). Note
that in this example transaction 126 was initiated by remote transaction 77 on another node. That means
transaction 77 is the root of the blocked transaction chain. See Blocked Transaction Monitoring.

Thread Monitor

The Thread Monitor allows the most fine-grained view into the current situation by listing all threads in the
system. Note that it is usually not necessary to drill into that level of detail. Threads contributing to a
transaction that is currently blocked are marked by a warning sign in the “Status” column. To get additional
information about the blocking situation, hover the mouse over the warning sign. See Thread Monitoring.

SAP HANA Troubleshooting and Performance Analysis Guide

76 PUBLIC Root Causes And Solutions
 Figure 22: Threads contributing to Blocked Transactions

Related Information

Load Monitoring [page 150]

Job Progress Monitoring [page 149]
Savepoint Performance [page 51]
Session Monitoring [page 148]
Blocked Transaction Monitoring [page 147]

3.8.1.2 Troubleshooting Blocked Transactions

When troubleshooting blocked transactions, it is helpful to differentiate between situations where only single
or a few transactions are blocked from the situation where a high percentage of all transactions are blocked.

3.8.1.2.1 Single or Few Transactions are Blocked

If you identified only a single or a few blocking transactions, there is likely an issue on application side.

A usual pattern is a flaw in the application coding that does not commit a write transaction. Such a transaction
will be a blocker for any other transaction that needs to access the same database object. To release the
situation you have to close the blocking transaction.

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 77
There are several possibilities to achieve this:

● Contact the Application User

The Session Monitor allows you to identify the user of the application. You can find this information in the
“Database User” column or, in case the application has its own user management (for example, SAP BW),
in the “Application User” column. Contact the user and ask him whether he can close the application.
● Contact the Application Developer
As a follow-up, the author of the application should be contacted whether such situations can be avoided
in the future by changing the application code.

3.8.1.2.1.1 Cancel the Session

If you are not able to contact the user to have them cancel the session, you can also cancel the session in the
context menu of the Session Monitor. The current transaction will be rolled back.

The session cancellation may take some time to succeed. If it takes longer than 30 seconds, consider this as a
bug and contact development support.

3.8.1.2.1.2 Kill the Client Appication

In case the session cancellation takes too long or does not complete at all, you can kill the client process that
opened the session. This will terminate the blocking transaction as well.

As a prerequisite, you have to have access to the client machine. The information needed for this task can be
retrieved from the Session Monitor.

See columns “Client IP” and “Client Process ID” to determine the host and process to be killed. Note that
killing the client application is safe from a database consistency standpoint, the current transaction will be
rolled back gracefully.

3.8.1.2.2 Many Transactions are Blocked

In the case that a large amount of transactions are blocked, the troubleshooting should take a slightly different
approach.

First you need to determine whether there is a single or few blocking transactions that block a large amount of
other transactions. For this, open the Blocked Transaction Monitor and check the amount of blocking
transactions. If you assess there is only a few blocking transactions, use the techniques described in Single of
Few Transactions are Blocked to resolve the situation.

If there are many transactions in a blocking state, you need to find out whether a specific access pattern
causes the situation. In case that multiple transactions try to access the same database objects with write
operations, they block each other. To check if this situation exists, open the Blocked Transaction Monitor and
analyze the “Waiting Schema Name”, “Waiting Object Name” and “Waiting Record Id” columns. If you find a
fair amount of blocking transactions that block many other transactions you need to investigate if the following
is possible:

SAP HANA Troubleshooting and Performance Analysis Guide

78 PUBLIC Root Causes And Solutions
● Change the client application(s) to avoid the access pattern
● If a background job is running that issues many write transactions (for example, a data load job):
Reschedule to a period with a low user load
● Partition tables that are accessed frequently to avoid clashes. See the SAP HANA Administration Guide for
more details on partitioning.

In case you cannot identify specific transactions or specific database objects that lead to transactions being
blocked, you have to assume a problem with the database itself or its configuration. One example is an issue
with long savepoint durations. See Savepoint Performance for troubleshooting such issues.

Related Information

Single or Few Transactions are Blocked [page 77]

Savepoint Performance [page 51]
SAP HANA Administration Guide

3.8.2 Troubleshooting Blocked Transaction Issues that

Occurred in the Past

Finding the root cause of blocked transaction situations that you have resolved is more difficult than
troubleshooting issues that are currently happening. Tools such as the Load Monitor, system views and the
SQL Plan cache are available to help you.

First use the Load Monitor to isolate the exact time frame where the issue happened. Using that information,
investigate what happened at this specific time frame. You should check the following monitoring and
StatisticServer views:

● _SYS_STATISTICS.HOST_BLOCKED_TRANSACTIONS: Analyze the columns

“WAITING_SCHEMA_NAME”, “WAITING_TABLE_NAME” and “WAITING_RECORD_ID” to identify the
database objects that lead to blocked transactions
● SYS.M_DELTA_MERGE_STATISTICS: The column “START_TIME” and “EXECUTION_TIME” provide you
with the information if there was a Delta Table Merge running A longer history can be found in the
StatisticServer table _SYS_STATISTICS.HOST_DELTA_MERGE_STATISTICS
● SYS.SAVEPOINTS: Check if a savepoint was written during the time period. A longer history can be found
in _SYS_STATISTICS.HOST_SAVEPOINTS

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 79
In addition the SAP HANA studio SQL Plan Cache monitor may be able to provide information about the
statements that were involved in the situation:

Figure 23: Plan Cache Monitor

Only check entries that have “TOTAL_LOCK_WAIT_COUNT” > 0. For those entries, compare the column
“MAX_CURSOR_DURATION” against “AVG_CURSOR_DURATION”. If there is a significant difference, there
was at least one situation where the transactions took much longer than average. This can be an indication
that it was involved in the situation.

3.8.3 Multiversion Concurrency Control (MVCC) Issues

In this section you will learn how to troubleshoot issues arising from MVCC.

Multiversion Concurrency Control (MVCC) is a concept that ensures transactional data consistency by
isolating transactions that are accessing the same data at the same time.

To do so, multiple versions of a record are kept in parallel. Issues with MVCC are usually caused by a high
number of active versions. Old versions of data records are no longer needed if they are no longer part of a
snapshot that can be seen by any running transaction. These versions are obsolete and need to be removed
from time to time to free up memory.

This process is called Garbage Collection (GC) or Version Consolidation. It can happen that a transaction is
blocking the garbage collection. The consequence is a high number of active versions and that can lead to
system slow-down or out of memory issues.

3.8.3.1 Row Store Tables

Garbage collection is triggered after a transaction is committed and also periodically (every hour by default).

A transaction that is currently committing can be identified in the Threads tab (see System Performance
Analysis). The Thread type will be “SqlExecutor” and the Thread method “commit”.

The periodic garbage collection can be identified by Thread Type” MVCCGarbageCollector”.

Note that the periodic garbage collection interval can be configured in the indexserver.ini file transaction
section with the parameter mvcc_aged_checker_timeout (up to SPS 10) or gc_interval (from SPS 11)..

SAP HANA Troubleshooting and Performance Analysis Guide

80 PUBLIC Root Causes And Solutions
Related Information

System Performance Analysis [page 144]

3.8.3.2 MVCC Problem Identification

There are a number of indicators of MVCC problems to check for.

Problems with high number of active versions can be identified by

● Users report an increase of response times

● The indexserver trace contains "There are too many un-collected versions. The transaction blocks the
garbage collection of HANA database."
● By checking “Active Versions” in the Load Monitor (see Performance tab)

Figure 24: Load Monitor Showing Garbage Collection

Transactions blocking garbage collection can originate from:

● Long-running or unclosed cursors

● Long-running transactions with isolation mode “serializable” or ”repeatable read”
● Hanging threads

In order to validate there is a problem with MVCC, check the number of active versions in the Row StoreMVCC
manager monitoring view. Note that in multihost environment, you have to check the master host.

select * from m_mvcc_tables where host='ld9989' and port='30003' and

(name='NUM_VERSIONS' or name='MAX_VERSIONS_PER_RECORD' or
name='TABLE_ID_OF_MAX_NUM_VERSIONS');

Figure 25: MVCC Information on a Healthy System

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 81
If the number of active versions (NUM_VERSIONS) is greater than eight million, it is considered a problem and
an overall slowdown of the system can be experienced. Similarly, if the maximum number of versions per
record (MAX_VERSIONS_PER_RECORD) exceeds 8,000,000, this should be treated as a problem and a
slowdown of accesses to a specific table is expected. Use TABLE_ID_OF_MAX_NUM_VERSIONS and join it
against the SYS.TABLES system view to determine the table which is having the problem.

Related Information

Performance Trace [page 180]

3.8.3.3 Analysis of MVCC Issues

You have to find which transactions are blocking the garbage collection and to which connection they are
related.

The following queries will return the transaction that may block the Garbage Collection. You have to check
both.

SELECT top 1 host, port, connection_id, transaction_id, update_transaction_id,

primary_transaction_id, transaction_type, isolation_level FROM M_TRANSACTIONS
WHERE
MIN_MVCC_SNAPSHOT_TIMESTAMP > 0 order by min_mvcc_snapshot_timestamp desc;

SELECT top 1 host, port, connection_id, transaction_id, update_transaction_id,

primary_transaction_id, transaction_type, isolation_level FROM M_TRANSACTIONS
WHERE
MIN_MVCC_SNAPSHOT_TIMESTAMP = (SELECT MIN(VALUE) FROM M_MVCC_TABLES WHERE NAME =
'MIN_SNAPSHOT_TS') order by min_mvcc_snapshot_timestamp desc;

Figure 26: User Transaction Possibly Blocking Garbage Collection

In case of a user transaction being the candidate (TRANSACTION_TYPE=’USER TRANSACTION’), you can
directly determine the connection ID the transaction belongs to (see an example in the Figure above).

Figure 27: External Transaction Possibly Blocking Garbage Collection

If the candidate’s transaction type is ‘EXTERNAL TRANSACTION’, use the following query to find out which
other transaction spawned the candidate and determine its connection ID.

SELECT t.connection_id AS "Kill this connection id",

t.transaction_id AS "Belonging to user transaction id",
e.transaction_id AS "To get rid of external transaction id"
FROM m_transactions t
JOIN m_transactions e ON
e.primary_transaction_id = t.transaction_id
AND e.volume_id = t.volume_id

SAP HANA Troubleshooting and Performance Analysis Guide

82 PUBLIC Root Causes And Solutions
 WHERE e.transaction_type = 'EXTERNAL TRANSACTION' and e.transaction_id = <GC
blocker transaction id>;

3.8.3.4 Solution of MVCC Issues

Solving MVCC issues is similar to solving blocked transaction issues. Use the following approaches in the given
order for transactions where you know the connection ID.

1. Contact the user to stop his activity

2. Cancel the statement/cancel internal transaction
3. Cancel connection
4. Kill the client application

Note
There is no guaranteed that these measures will stop a transaction which blocks the garbage collection. If
that is the case, contact development support immediately to get further help

Related Information

Resolving CPU Related Issues [page 38]

3.8.4 Version Garbage Collection Issues

Alerts 73, 74, and 75 help you to identify and resolve version space overflow issues.

Context

The following steps allow you to check whether or not the issue you have is related to a “version space
overflow/skew” situation.

Procedure

1. The first step is to check the alerts.

The following alerts are indicators that could be the problem:
○ Alert 73 Rowstore version space overflow, determines the overflow ratio of the rowstore version
space.

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 83
 ○ Alert 74 Metadata version space overflow, determines the overflow ratio of the metadata version
space.
○ Alert 75 Rowstore version space skew, determines whether the rowstore version chain is too long.

If you use extended storage, a version space overflow may manifest as an out of space error for the delta
dbspace, and will not trigger these alerts. See Out of Space Errors for Delta Dbspace in SAP HANA
Dynamic Tiering: Administration Guide.
2. Identify the statement and connection blocking garbage collection.
a. Identify the connection or transaction that is blocking version garbage collection.
The System Information tab of SAP HANA studio provides a set of tabular views to display blocking
connections based on pre-defined SQL queries.

○ The view MVCC Blocker Connection shows connections that may be blocking garbage collection.
○ The view MVCC Blocker Statement shows statements that may be blocking garbage collection.
○ The view MVCC Blocker Transaction shows transactions that may be blocking garbage collection.

In the first row of the table shown below if there is a "global" garbage collection blocker whose
IDLE_TIME > 3600 seconds (1 hour) then investigate what the statement is doing and take the
necessary corrective action.

In the second row, a table level garbage collection blocker whose IDLE_TIME > 3600 seconds (1 hour)
can be seem. In this case you can query the following monitoring view to check how many versions the
related table TEST2 has.

SELECT * FROM M_TABLE_SNAPSHOTS WHERE TABLE_NAME = 'TEST2' AND

START_MVCC_TIMESTAMP = 142024

If the results of the query shows that VERSION_COUNT > 1 million, the blocking statement can cause
a performance drop for table updates. In this case, investigate what the statement is doing and take
the necessary corrective action. However, if VERSION_COUNT is less than 10,000 its impact on
performance is negligible.

Note that the table only shows selected columns from the results of running MVCC Blocker Statement
for the purposes of this example. If there is no blocker in MVCC Blocker Statement, then use MVCC
Blocker Transaction. The blocker would be likely one of internal/external/serializable transaction.

Table 6:
Type IDLE_TIME_SEC CONNEC­ START_MVCC_ STATE­ INFO
TION_ID TIMESTAMP MENT_STRING

GLOBAL 4000 200285 142113 SELECT * FROM Check VER­

TEST1 SION_COUNT in
M_MVCC_OVER­
VIEW

TABLE 5000 200375 142024 SELECT * FROM Check VER­

TEST2 ORDER SION_COUNT in
BY A M_MVCC_TA­
$”ŁLyf£N˚BžBért\�Ã
SHOTS

SAP HANA Troubleshooting and Performance Analysis Guide

84 PUBLIC Root Causes And Solutions
 Type IDLE_TIME_SEC CONNEC­ START_MVCC_ STATE­ INFO
TION_ID TIMESTAMP MENT_STRING

TABLE 100 200478 142029 SELECT * FROM Check VER­

TEST0 ORDER SION_COUNT in
BY A M_MVCC_TA­
¾ÌajOçˆÊ×aßïu<0®
jØ
SHOTS

3. Kill the statement and/or connection that is blocking garbage collection.

a. Kill the connection

Check MVCC Blocker Statement to find out which statement is blocking version garbage collection
and investigate what the blocker is doing. If the TYPE column is GLOBAL, then it is a global version
garbage collection blocker. The column IDLE_TIME_SEC will help your decision. If TYPE is TABLE, it’s
a blocker of the specific table. With M_TABLE_SNAPSHOTS, you can find out the number of versions
for each table.

To kill the blocking connection use ALTER SYSTEM DISCONNECT SESSION 'CONNECTION_ID'
b. Kill the transaction, if killing the connection does not succeed or CONNECTION_ID does not exist..

Check MVCC Blocker Transaction to find out which transaction blocks global version garbage
collection and investigate what the blocker is doing. The column LIFE_TIME_SEC column will help your
decision.

To cancel a transaction that you have identified as the cause of the blockage open the Emergency
Information tab in the SAP HANA studio, right click on the transaction in the tabular view and choose
Cancel Transaction <TRANSACTION ID>.

Note
Be aware that TRANSACTION ID is only unique on a specific node and therefore the ID might be
reused. So make sure you are connected to the right node before you cancel the transaction.

You can find a version garbage collection blocker with thread info using these queries.

MVCC Blocker Statement

WITH STORED_MVCC_BLOCKER_STATEMENTS AS (SELECT

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 85
 CASE
WHEN A.START_MVCC_TIMESTAMP = B.MIN_MVCC_SNAPSHOT_TIMESTAMP THEN
'GLOBAL'
ELSE 'TABLE'
END TYPE,
CASE
WHEN A.START_MVCC_TIMESTAMP = B.MIN_MVCC_SNAPSHOT_TIMESTAMP THEN
NULL
ELSE 'check M_TABLE_SNAPSHOTS for details'
END INFO,
SECONDS_BETWEEN(A.LAST_ACTION_TIME,
CURRENT_TIMESTAMP) IDLE_TIME_SEC,
A.*
FROM
M_ACTIVE_STATEMENTS AS A
JOIN
M_MVCC_OVERVIEW AS B
ON A.HOST = B.HOST
AND A.PORT = B.PORT
WHERE
A.START_MVCC_TIMESTAMP > 0
AND A.STATEMENT_STATUS <> 'NONE'
AND A.START_MVCC_TIMESTAMP <= B.MIN_MVCC_SNAPSHOT_TIMESTAMP
ORDER BY
IDLE_TIME_SEC DESC) SELECT
*
FROM
STORED_MVCC_BLOCKER_STATEMENTS AS A LEFT OUTER JOIN
M_SERVICE_THREADS AS B
WHERE
A.CONNECTION_ID = B.CONNECTION_ID;

MVCC Blocker Transaction

WITH STORED_MVCC_BLOCKER_TRANSACTIONS AS (SELECT

SECONDS_BETWEEN(A.START_TIME,
CURRENT_TIMESTAMP) LIFE_TIME_SEC,
A.*
FROM
M_TRANSACTIONS AS A
JOIN
M_MVCC_OVERVIEW AS B
ON A.HOST = B.HOST
AND A.PORT = B.PORT
WHERE
A.MIN_MVCC_SNAPSHOT_TIMESTAMP > 0
AND A.MIN_MVCC_SNAPSHOT_TIMESTAMP = B.MIN_MVCC_SNAPSHOT_TIMESTAMP
ORDER BY
LIFE_TIME_SEC DESC) SELECT
*
FROM
STORED_MVCC_BLOCKER_TRANSACTIONS AS A LEFT OUTER JOIN
M_SERVICE_THREADS AS B
WHERE
A.CONNECTION_ID = B.CONNECTION_ID;

MVCC Blocker Connection

WITH STORED_MVCC_BLOCKER_CONNECTIONS AS
(SELECT *
FROM
M_CONNECTIONS
WHERE
CONNECTION_ID IN (SELECT
CONNECTION_ID
FROM

SAP HANA Troubleshooting and Performance Analysis Guide

86 PUBLIC Root Causes And Solutions
 M_TRANSACTIONS AS A
JOIN
M_MVCC_OVERVIEW AS B
ON A.HOST = B.HOST
AND A.PORT = B.PORT
WHERE
A.MIN_MVCC_SNAPSHOT_TIMESTAMP > 0
AND A.MIN_MVCC_SNAPSHOT_TIMESTAMP <=
B.MIN_MVCC_SNAPSHOT_TIMESTAMP))
SELECT *
FROM
STORED_MVCC_BLOCKER_CONNECTIONS AS A LEFT OUTER JOIN
M_SERVICE_THREADS AS B
WHERE
A.CONNECTION_ID = B.CONNECTION_ID;

3.9 Statement Performance Analysis

This section gives an overview of issues and solutions concerning SQL statement performance.

3.9.1 SQL Statement Optimization

This section provides an overview of tools, traces and SAP HANA studio areas that can be used to identify
critical SQL statements.

SQL statements that are not executed efficiently can cause local and system-wide problems. The most critical
are the following areas:

● A long runtime can result in delays of the business activities

● A high CPU consumption can lead to system-wide CPU bottlenecks
● High memory consumption can be responsible for out-of-memory situations and performance penalties
due to unload of tables from memory

SQL statements consuming significant resources are called expensive SQL statements.

Identification of Critical SQL Statements

A key step in identifying the source of poor performance is to understand how much time is spent in the SAP
HANA engine for query execution. By analyzing SQL statements and calculating their response times, you can
better understand how the statements affect application and system performance.

Before you are able to analyze and optimize a SQL statement you have to identify the critical SQL statements.
We can distinguish between the following scenarios:

● SQL statements that have caused problems in the past

● SQL statements that are currently causing problems

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 87
3.9.2 SQL Statements Responsible for Past Problems

Sometimes it is not possible to identify a critical SQL statement during runtime. In this case you can use the
following approaches to identify one or several SQL statements that can have contributed to the problem.

You can identify a SQL statement either by its SQL text (“statement string”) or by the related statement hash
that is uniquely linked to an individual SQL text. The mapping of the statement hash to the actual SQL text is
described later.

● To determine SQL statements with a particularly high runtime you can check for the top SQL statements
in terms of TOTAL_EXECUTION_TIME in the SQL plan cache in SAP HANA Studio on the Performance
SQL Plan Cache tab.

Figure 28: SQL Statements With High Runtimes

● To determine the top SQL statements that were executed during a dedicated time frame in the past, you
can check the SQL plan cache history (HOST_SQL_PLAN_CACHE). You can use the SQL statement:
“HANA_SQL_SQLCache_History” available from SAP Note 1969700 – SQL Statement Collection for SAP
HANA in order to check for top SQL statements during a specific time frame:
You have to specify a proper BEGIN_TIME / END_TIME interval and typically use ORDER_BY =
‘ELAPSED’, so that the SQL statements with the highest elapsed time from SAP HANA are returned.

Figure 29: SQL Statements With Highest Elapsed Time

● The thread sample history (tables M_SERVICE_THREAD_SAMPLES,
HOST_SERVICE_THREAD_SAMPLES), if available, can also be used to determine the top SQL statements.
You can use the SQL statement: “HANA_Threads_ThreadSamples_FilterAndAggregation” available from
SAP Note 1969700 – SQL Statement Collection for SAP HANA in order to check.
You have to specify a proper BEGIN_TIME / END_TIME interval and use AGGREGATE_BY =
‘STATEMENT_HASH’ to identify the top SQL statements during the time frame.

SAP HANA Troubleshooting and Performance Analysis Guide

88 PUBLIC Root Causes And Solutions
 Figure 30: SQL Example Output

In this case the SQL statement with hash 51f62795010e922370bf897325148783 is executed most often
and so the analysis should be started with it. Often you need to have a look at some more SQL statements,
for example the statements related to the next statement hashes fc7de6d7b8942251ee52a5d4e0af728f
and 1f8299f6cb5099095ea71882f84e2cd4.
● In cases where the M_SERVICE_THREAD_SAMPLES / HOST_SERVICE_THREAD_SAMPLES information
is not usable you can use the thrloop.sh script to regularly collect thread samples as described in SAP
Note 1989031 – Scheduling of Shell script “thrloop.sh”
● In case of an out-of-memory (OOM) situation you can determine potentially responsible SQL statements
by analyzing the OOM dump file(s) as described in SAP Note1984422 – Analysis of HANA Out-of-memory
(OOM) Dumps.
● SAP HANA Alert 39 (“Long running statements”) reports long running SQL statements and records them
in the table _SYS_STATISTICS.HOST_LONG_RUNNING_STATEMENTS. Check the contents of this table
to determine details of the SQL statements that caused the alert.

Related Information

SAP Note 1969700

SAP Note 1989031
SAP Note 1984422

3.9.3 SQL Statements Responsible for Current Problems

If problems like high memory consumption, high CPU consumption or a high duration of individual database
requests are currently happening, you can determine the active SQL statements with the help of SAP HANA
studio.

Check for the currently running SQL statements in SAP HANA studio on the Performance Threads tab.

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 89
 Figure 31: Running SQL Statements in Threads Detail

3.9.4 SQL Statements Reported by Traces

Additional traces can be activated to help identify critical SQL statements.

The following standard traces exist to identify critical SQL statements. Activating traces in addition to the
normally available performance information can be useful for the following reasons:

● Capturing of more detailed information

● Determination of SQL statements that are not obvious on a global level, but important for business critical
applications.

The following traces are available to determine SQL statements:

● SQL trace
○ Capturing of performance data for all SQL statements
○ Filter conditions can be used to limit the trace activities

● Expensive statements trace

○ Captures all SQL statements with a runtime beyond a defined threshold

Further details can be found in Tools and Tracing.

Related Information

Tools and Tracing [page 144]

3.9.5 Analysis of Critical SQL Statements

When you have identified the SQL text and the related statement hash based on the tools and traces, you can
collect more information about the SQL statement in order to identify the root cause and optimize the SQL
statement. The available analysis approaches are described here.

From a technical perspective, analyzing query plans allows you to identify long running steps, understand how
much data is processed by the operators, and see whether data is processed in parallel. However, if you
understand the idea and purpose behind the query, you can also analyze query plans from a logical
perspective and consider the following questions to gain the insight you need:

● Does SAP HANA read data from multiple tables when only one is required?
● Does SAP HANA read all records when only one is required?

SAP HANA Troubleshooting and Performance Analysis Guide

90 PUBLIC Root Causes And Solutions
● Does SAP HANA start with one table even though another table has a much smaller result set?

3.9.6 SQL Plan Cache Analysis

The SAP HANA SQL Plan Cache can be evaluated in detail for a particular statement hash.

You can use the SQL statement: “HANA_SQL_StatementHash_KeyFigures” available in SAP Note 1969700 –
SQL Statement Collection for SAP HANA in order to check for the SQL Plan Cache details of a specific SQL
statement (the related STATEMENT_HASH has to be maintained as input parameter).

Figure 32: SQL PLAN Cache Example Output

Alternatively you can check the SAP HANA view M_SQL_PLAN_CACHE or the Performance SQL Plan
Cache tab in SAP HANA studio.

The historic execution details for a particular SQL statement can be determined with the SQL statement:
“HANA_SQL_SQLCache” included with SAP Note 1969700. Also here the appropriate STATEMENT_HASH has
to be specified as input parameter.

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 91
 Figure 33: Statement Hash Example output:

Based on the results of this evaluation you can distinguish the following situations:

● If the value for Executions is unexpectedly high, further analysis should be done on the application side in
order to check if it is possible to reduce the number of executions.
● If the value for Records is unexpectedly high, further analysis should be done on the application side in
order to check if it is possible to reduce the number of selected records.
● If the value for Cursor duration is very high and at the same time significantly higher than the value for
Execution time, you have to check which processing steps are executed on the application side between
the individual fetches. A high value for Cursor duration can negatively impact the database in general
because open changes may impact the MVCC mechanism.
● If the value for Preparation time is responsible for a significant part of the Execution time value you have to
focus on optimizing the parsing (for example, sufficient SQL plan cache size, reuse of already parsed SQL
statements).
● If Execution time is much higher than expected (that can be based on the statement complexity and the
number of processed rows), the SQL statement has to be checked more in detail on technical layer to
understand the reasons for the high runtime. See section “Query Plan Analysis” for more information.

Related Information

SAP Note 1969700

Query Plan Analysis [page 157]
Example: Reading the SQL Plan Cache [page 93]

SAP HANA Troubleshooting and Performance Analysis Guide

92 PUBLIC Root Causes And Solutions
3.9.7 Example: Reading the SQL Plan Cache

These examples aim to show you how to gain useful insights by analyzing the SQL plan cache.

Execution in a Single-Host System

This example aims to show you how to interpret information about execution time. The following table is
sorted by TOTAL_EXECUTION_TIME.

Table 7:
USER_ STATMENT_ STRING TOTAL_ EXECUTION_ AVG_ EXECUTION_ EXECUTION_ COUNT
NAME TIME TIME

SYSTEM SELECT "REQUEST" , 774,367,833 181,266 4,272

"DATAPAKID" ,
"PARTNO" , "RECORD" ,
"CALDAY" , ...

SYSTEM SELECT * FROM "/BIC/ 726,672,877 60,556,073 12

AZDSTGODO40" WHERE
"SID" = ?

SYSTEM SELECT "JOBNAME" , 473,620,452 22,987 20,604

"JOBCOUNT" ,
"JOBGROUP" ,
"INTREPORT" ,
"STEPCOUNT" ...

– <Further 6832 records> – – –

You could read these top 3 results as follows:

● Statement 1 takes the longest time overall but it is also executed frequently.
● Statement 2 is not executed very frequently but has the second highest total execution time. Why is this
simple SQL taking so long? Does it have problems processing?
● The execution times for statement 3 are fine for one-off execution, but it runs too frequently, over 20,000
times. Why? Is there a problem in application code?

Sorting by AVG_EXECUTION_TIME or EXECUTION_COUNT provides a different angle on your analysis.

The following example aims to show you how to interpret information about locking situations. The information
in columns TOTAL_LOCK_WAIT_COUNT and TOTAL_LOCK_WAIT_DURATION lets us know which statement
is waiting for others and how much time it takes.

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 93
Table 8:
USER STATEMENT_STRING TOTAL_LOCK_ TOTAL_LOCK_ TOTAL_ EXECU­
WAIT_COUNT WAIT_DURATION TION_TIME

SYSTEM SELECT 11,549,961 210,142,307,207 676,473

"FROMNUMBER","TONUMBER"
,"NRLEVEL" FROM
"NRIV" ... FOR UPDATE

SYSTEM UPDATE "NRIV" SET 0 0 3,706,184

"NRLEVEL" = ? WHERE
"CLIENT" = '000' ...
AND "TOYEAR" = '0000'

SYSTEM SELECT "DIMID" FROM 0 0 460,991

"/BIC/DZDSTGCUBE4"
WHERE "/B49/S_VERSION"
= ?

Here, it is clear that the first statement is waiting almost all the time. Known as pessimistic/optimistic locking,
the SELECT...FOR UPDATE code locks the resulting columns and may be replaced by a non-locking variant,
which can result in poor performance. If the application is critical, it may be necessary to revise the
SELECT...FOR UPDATE code for better resource utilization and performance.

Execution in a Distributed System

In distributed SAP HANA systems, tables and table partitions are located on multiple hosts. The execution of
requests received from database clients may potentially have to be executed on multiple hosts, depending on
where the requested data is located. The following example illustrates statement routing and how, if it is not
enabled, requests from the database client are executed on the contacted index server (in this case the
master index server) and the required data is fetched from the index server on the relevant host(s). However,
if statement routing is enabled, after initial query compilation, request execution is routed directly to the host
on which the required data is located.
Figure 34: Distributed Execution with Statement Routing Off and On

SAP HANA Troubleshooting and Performance Analysis Guide

94 PUBLIC Root Causes And Solutions
Execution times should be better with statement routing enabled. You can use the SQL plan cache to compare
the execution statistics of statements with statement routing enabled and disabled and thus confirm the
effect.

Statement routing is controlled by the client_distribution_mode parameter in the indexserver.ini

file. It is enabled by default (value=statement).

The following SQL plan cache examples show the execution times of sample statements based on the scenario
illustrated above.

Note
The column IS_DISTRIBUTED_EXECUTION indicates whether or not statement execution takes place on
more than one host.

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 95
The TOTAL_EXECUTION_TIME for a statement is the sum of execution times on all hosts, therefore:

Table 9:
Statement Request Path Total Execution Time

UPSERT "RSBMONMESS_DTP" seltera12 = execution time on seltera12

( "MSGNO", "MSGTY", "MSGID", ...

SELECT * FROM "/BI0/SIOBJNM" seltera12 > selbld13 = execution time on seltera12 + execu­
WHERE "IOBJNM" = ? tion time on selbld13

SELECT * FROM "/B49/SCUSTOMER" seltera12 > seltera13 > selbld16 = execution time on seltera12 + execu­
WHERE "/B49/S_CUSTOMER" = ? tion time on selbld16

Table 10:
Statement Request Path Total Execution Time

UPSERT "RSBMONMESS_DTP" seltera12 = execution time on seltera12

( "MSGNO", "MSGTY", "MSGID", ...

SELECT * FROM "/BI0/SIOBJNM" selbld13 = execution time on selbld13

WHERE "IOBJNM" = ?

SELECT * FROM "/B49/SCUSTOMER" selbld16 = execution time on selbld16

WHERE "/B49/S_CUSTOMER" = ?

3.9.8 Detailed Statement Analysis

When you have identified a critical SQL statement and identified its overall key figures from the SQL plan
cache analysis you can have a closer look at the actual runtime behavior of the SQL statement.

The following tools can be used for a more detailed analysis:

● Plan Explanation
○ Creation of an execution plan
● Plan Visualizer
○ Detailed graphical execution plan
○ Temporal breakdown (timeline) available
● QO Trace

SAP HANA Troubleshooting and Performance Analysis Guide

96 PUBLIC Root Causes And Solutions
 ○ Query optimizer trace
○ Advanced tool that can be useful to understand the decisions of the query optimizer
○ Particularly helpful to understand column searches
● JE Trace
○ Join evaluation trace
○ Advanced tool to analyze table join operations
● Performance trace
○ Low level recording of key performance indicators for individual SQL statement processing steps
○ Advanced tool that should only be used in collaboration with SAP support
● Kernel profiler
○ Sample based profiling of SAP HANA process activities
○ Advanced tool that should only be used in collaboration with SAP support

All these tools are described in Tools and Tracing.

Related Information

Tools and Tracing [page 144]

3.9.9 Optimization of Critical SQL Statements

You can improve the general performance of the SAP HANA database by implementing various best practices,
design principles, available features, and add-ons.

To enhance the performance of the SAP HANA database, we recommend you do the following:

● Optimize outlier queries

Queries that sometimes take much longer than expected can be caused by query-external factors (for
example, resource bottlenecks) that have to be determined and eliminated.
● Check data manipulation commands (DML)
DML operations like INSERT, UPDATE and DELETE can be impacted by lock waits.
● Improve SQL query design
You can significantly improve the performance of your SQL queries by knowing how the SAP HANA
database and SAP HANA engines process queries and adapting your queries accordingly.
● Create indexes for any non-primary key columns that are often queried.
SAP HANA automatically creates indexes for primary key columns; however, if you need indexes for non-
primary key columns, you must create them manually.
● Create virtual star schemas on SAP HANA data by joining attribute views to tables that contain measures
within the definition of an analytic view
By using analytic views, SAP HANA can automatically recognize star queries and enable the performance
benefits of using star schemas, such as reducing dependencies for query processing.
● Develop procedures to embed data-intensive application logic into the database.
With procedures, no large data transfers to the application are required and you can use performance-
enhancing features such as parallel execution. Procedures are used when other modeling objects, such as
analytic or attribute views, are not sufficient.

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 97
 ○ If you use SQLscript to create procedures, follow the best practices for using SQLScript.
○ For statistical computing, create procedures using the open source language R.
● Download and install the available application function libraries, such as Predictive Analysis Library (PAL)
and Business Function Library (BFL).
Application functions are like database procedures written in C++ and called from outside to perform data
intensive and complex operations.
● Scale SAP HANA to improve performance.
SAP HANA's performance is derived from its efficient, parallelized approach. The more computation cores
your SAP HANA server has, the better overall system performance is.

Note
With SAP HANA, you do not need to perform any tuning to achieve high performance. In general, the SAP
HANA default settings should be sufficient in almost any application scenario. Any modifications to the
predefined system parameters should only be done after receiving explicit instruction from SAP Support.

3.9.10 Outlier Queries

Outlier queries are database statements that take much more time than usual and expected. This usually
happens because extra work has to be performed during execution.

Besides heavy load on the machine by non-SAP HANA processes (which should not be the case on production
systems) SAP HANA itself can be under heavy load. Reasons include:

● Many users are connected and issue a large amount of statements

● Extraordinary expensive statements are executed
● Background tasks are running

SAP HANA Troubleshooting and Performance Analysis Guide

98 PUBLIC Root Causes And Solutions
Use the Load Monitor to determine the number of statements issued and the indexserver CPU usage while the
slow statement execution was perceived ( See the figure Load Monitor in SAP HANA Studio, the lower line
(red) is the CPU consumption in percent (%), the upper line (orange) is the SQL throughput / s):

Figure 35: Load Monitor in SAP HANA Studio

You can see that during the period in the red rectangle both CPU consumption and SQL throughput
decreased. During that time frame you would look for something that consumed a lot of resources or blocked
the statements (locking); just after 15:35 you see that the CPU consumption increases while the SQL
throughput decreases. Here, a possible case would be a change in usage: instead of many small, fast SQL
statements the workload changed to a few "heavy" (complicated calculation requiring many CPU cycles) SQL
statements.

If there was a high statement load in the same period when you experienced the slow execution the root cause
is likely a lack of resources. To resolve the situation consider restricting the number of users on SAP HANA,
upgrading the hardware, or get in touch with SAP Support if scalability can be improved in this case.

If you did not experience a high statement load during the time frame of the problem, check for background
activities:

● Delta Merges: Use Load Monitor Column Store Merge Requests and the monitoring view
M_DELTA_MERGE_STATISTICS to check if delta merges happened. In that case try to improve the delta
merge strategy to prevent merges happening in phases where users are disturbed (see the SAP HANA
Administration Guide for details).
● Column Unloads: See Load Monitor Column Store Column Unloads and the Monitoring View
M_CS_UNLOADS to look for signs of column unloads. If a column used in the problematic statement had
to be loaded before execution, the execution itself will take significantly longer.
● Savepoints: Savepoints consume resources and write-lock the database during their critical phase. Check
M_SAVEPOINTS and look for savepoints during the time frame of the problem. If a savepoint slowed down
your execution, the chance of having the same problem again is very low. If it happens often, please
contact SAP Support.

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 99
Related Information

SAP HANA Administration Guide

M_DELTA_MERGE_STATISTICS
M_CS_UNLOADS
M_SAVEPOINTS

3.9.11 Data Manipulation Language (DML) Statements

Data Manipulation Language (DML) statements are often slowed down by lock-wait situations.

Check under Performance SQL Plan Cache and the view M_SQL_PLAN_CACHE to determine such
issues:

Figure 36: Plan Cache Monitor

Note
Only check entries that have TOTAL_LOCK_WAIT_COUNT greater than 0. For those entries, compare the
column MAX_CURSOR_DURATION against AVG_CURSOR_DURATION. If there is a significant difference,
there was at least one situation where the transactions took much longer than average.

See Transactional Problems for information on how to deal with such issues.

Related Information

M_SQL_PLAN_CACHE
Transactional Problems [page 74]

SAP HANA Troubleshooting and Performance Analysis Guide

100 PUBLIC Root Causes And Solutions
3.9.12 SQL Query Tuning Tips

Performance of SQL queries can be improved significantly by knowing how the SAP HANA database and SAP
HANA engines process queries and adapting queries accordingly.

As a general guideline for improving SQL query performance, we recommend avoiding operations that are not
natively supported by the various SAP HANA engines since they can significantly increase the time required to
process queries.

Please note that following specific recommendations may help to improve the performance of SQL queries
involving column tables.

Caution
Throughout this section, adding generated columns will be mentioned as a possible workaround to improve
performance of the query. However, it should be noted that adding generated columns improves query
performance at the expense of increased memory consumption, and increased insertion and update cost.
You should be aware of this trade-off before deciding to add generated columns.

3.9.12.1 Avoid Implicit Type Casting in Queries

You can avoid implicit type casting by instead using explicit type casting or by adding generated columns.

The system can generate type castings implicitly even if the user did not explicitly write type casting operation.
For example, if there is a comparison between a VARCHAR value and a DATE value, the system generates an
implicit type casting operation that casts the VARCHAR value into a DATE value. Implicit type casting is done
from lower precedence type to higher precedence type. You can find the type precedence rule in the SAP
HANA SQL and System Views Referencel. If two columns are frequently compared by queries, it is better to
enforce that the two columns have the same data type from the beginning.

One way to avoid the cost of implicit type casting is by using explicit type casting on an inexpensive part of the
query. For instance, if a VARCHAR column is compared with a DATE value and the user knows that casting the
DATE value into a VARCHAR value produces what the user wants, it is recommended to cast the DATE value
into a VARCHAR value. if the VARCHAR column contains only date values in the form of 'YYYYMMDD', it could
be compared with a string generated from a DATE value in the form of 'YYYYMMDD'.

In the example below, date_string is a VARCHAR column. Note that comparison result between strings is
different from that between dates in general and identical only in some cases.

Table 11:
Problematic Query
SELECT * FROM T WHERE date_string <
CURRENT_DATE;

Workaround
SELECT * FROM T WHERE date_string <
TO_VARCHAR(CURRENT_DATE, 'YYYYMMDD');

If there is no way to avoid implicit type casting, one way to avoid implicit type casting entirely is to add
generated columns.

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 101
In the example below, user can find '1', '1.0' and '1.00' stored in a VARCHAR column using the revised query
and avoid implicit casting.

Table 12:
Problematic Query
SELECT * FROM T WHERE varchar_value =
1

Workaround
ALTER TABLE T ADD (num_value DECIMAL
GENERATED ALWAYS AS varchar_value);
SELECT * FROM T WHERE num_value = 1;

3.9.12.2 Avoid Inefficient Predicates in JOIN

This section will list out predicate conditions that are not natively supported by column engine.

Depending on the condition, intermediate results will be materialized and consumed by the row engine or
column engine. It is always a good practice to try to avoid intermediate result materialization since
materialization can be costly if results are large and have significant impact on the performance of the query.

3.9.12.2.1 Non-equijoin Predicate in Outer Join

The column engine natively support an outer join with a join predicate of equality and join predicates
connected by OR.

The column engine does not natively support an outer join with join predicates other than equality condition
(that is. non-equijoin) and join predicates connected by OR (even if each predicate is an equality condition).
Also, if equijoin predicates are connected by AND with non-equijoin predicates, it is processed in the same way
as the case with only non-equijoin predicates.

When such join predicates are used, the row engine executes the join operation using the appropriate join
algorithm (nested-loop, range, hashed-range) after materializing the intermediate results from both children.

An example of rewriting a non-equijoin predicate into an equijoin predicate is shown below. In the example, M
is a table containing first and last dates of months of interest.

Table 13:
Problematic Query
SELECT M.year, M.month,
SUM(T.ship_amount)
FROM T LEFT OUTER JOIN M ON
T.ship_date BETWEEN M.first_date AND
M.last_date
GROUP BY M.year, M.month;

SAP HANA Troubleshooting and Performance Analysis Guide

102 PUBLIC Root Causes And Solutions
 Workaround
SELECT M.year, M.month,
SUM(T.ship_amount)
FROM T LEFT OUTER JOIN M ON
EXTRACT(YEAR FROM T.ship_date) =
M.year AND EXTRACT(MONTH FROM
T.ship_date) = M.month
GROUP BY M.year, M.month;

3.9.12.2.2 Calculated Column Predicates

Intermediate results from calculations that are not natively supported by column engine are materialized but
can be avoided using generated columns.

As of SPS10, most of calculations are natively supported by column engine. But for a calculation that is not
natively supported by column engine, the intermediate results from both children are materialized and
consumed by row engine. One way to avoid calculation on query processing is to add generated columns. An
example of avoiding calculated join columns using generated columns is shown below. In the example, M is a
table containing first and last dates of months of interest.

Table 14:
Problematic Query
SELECT * FROM T JOIN M ON
NEXT_DAY(T.ship_date) = M.last_date;

Workaround
ALTER TABLE T ADD (next_day DATE
GENERATED ALWAYS AS
ADD_DAYS(ship_date, 1));
SELECT * FROM T JOIN M ON T.next_day
= M.last_date;

Calculation with parameters are not natively supported by column engine. For such queries, intermediate
results will be materialized and will be consumed by column engine. For such cases, there can be performance
impact if the amount of materialization is large.

Also, calculation involving multiple tables are not natively supported by column engine. This will result in
materialization of the intermediate results which will be consumed by row engine. One way to avoid such
calculation is to maintain needed columns in different tables by changing schema design.

We recommend trying to rewrite the query to avoid such conditions, if possible.

Table 15:
Predicate with parameter is not natively supported
SELECT * FROM T JOIN M ON
T.ship_amount + ? = M.target_amount;

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 103
Calculation involving multiple tables is not natively sup­
ported
SELECT * FROM T JOIN M ON
IFNULL(T.ship_amount,
M.pre_order_amount) = M.target_amount;

If applicable, the following hints can be tried to optimize performance for predicates involving calculated
columns.

WITH HINT(CS_EXPR_JOIN) --prefer column engine calulation

WITH HINT(NO_CS_EXPR_JOIN) --avoid column engine calculation

3.9.12.2.3 Cyclic Join

The column engine does not natively support join trees that have cycles in join edges if an outer join is involved
in the cycle (also known as cyclic outer join)

If there is such a cycle involving an outer join, the result from a child of the join that completes the cycle is
materialized to break the cycle.

Column engine does supports cyclic inner join natively, but it is better to avoid it because its performance is
inferior to acyclic inner join.

One way of breaking cycle is to maintain needed columns in different tables by changing schema design. For
the acyclic join in the example below, nation column of supplier table is moved to a line-item table.

Table 16:
Cyclic join
SELECT * FROM supplier S, customer C,
lineitem L
WHERE L.supp_key = S.key AND
L.cust_key = C.key AND S.nation =
C.nation;

Acyclic join
SELECT * FROM supplier S, customer C,
lineitem L
WHERE L.supp_key = S.key AND
L.cust_key = C.key AND L.supp_nation
= C.nation;

SQL optimizer selects cyclic inner join based on cost and it is sometime worthwhile to try hints to guide
optimizer to break cyclic join into two column searches and vice versa.

The following hints can be tried to optimize performance for cyclic inner join.

WITH HINT(CYCLIC_JOIN) --prefer cyclic join

WITH HINT(NO_CYCLIC_JOIN ) --break cyclic join

SAP HANA Troubleshooting and Performance Analysis Guide

104 PUBLIC Root Causes And Solutions
3.9.12.2.4 Filter Predicate Inside an Outer Join

The column engine does not natively support filter predicates inside outer join predicates.

Filter predicates over the right child of left outer join and filter predicates over the left child of a right outer join
are exceptions because moving those predicates below a join produces the same results.

Table 17:
Filter that is natively supported
SELECT * FROM T LEFT JOIN S ON T.a =
S.a AND S.b = 1; (equivalent to
SELECT * FROM T LEFT JOIN (SELECT *
FROM S WHERE S.b = 1) S ON T.a = S.a)

Filter that is not natively supported

SELECT * FROM T LEFT JOIN S ON T.a =
S.a AND T.b = 1;

If filter predicates are used inside outer join predicates and they cannot be moved below join, row engine
executes the join operation after materializing intermediate results from both children.

An example of rewriting such filter predicates into equijoin predicates using a generated column is shown
below.

Table 18:
Problematic query
SELECT * FROM T LEFT JOIN S ON T.a =
S.a AND T.b = 1;

Workaround
ALTER TABLE S ADD (one INTEGER
GENERATED ALWAYS AS 1);
SELECT * FROM T LEFT JOIN S ON T.a =
S.a AND T.b = S.one;

3.9.12.2.5 Subquery Filter Predicate Accessing Multiple

Tables Inside Outer Join

Filter predicates over multiple tables are not natively supported by column engine if they are under an outer
join.

If such a filter predicate exists, the result from the child including the predicate is materialized before
executing the join. Filter predicates over the left child of a left outer join and filter predicates over the right child
of a right outer join are exceptions. Moving those predicates upwards in the outer join produces the same
results and such a move is automatically done by the SQL optimizer. Below is an example of the filter predicate
that triggers materialization of intermediate result. One way of avoiding such materialization in the example
would be maintaining priority column in the lineitem table instead of orders table.

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 105
Table 19:
Problematic query
SELECT * FROM customer C LEFT OUTER
JOIN (
SELECT * FROM orders O JOIN
lineitem L ON O.order_key =
L.order_key
WHERE L.shipmode = 'AIR' OR
O.priority = 'URGENT') ON C.cust_key
= L.cust_key;

Workaround
SELECT * FROM customer C LEFT OUTER
JOIN (
SELECT * FROM lineitem L WHERE
L.shipmode = 'AIR' OR L.priority =
'URGENT') ON C.cust_key = L.cust_key;

3.9.12.2.6 Subquery Projection Of Constant Or Calculated

Value Below Outer Join

The column engine does not natively support constant or calculated value projection below outer joins.

If constant or calculated value projection exists, the result from the child including the projection is
materialized before executing the join.

Constant or calculated value projection over left child of left outer join or right child of right outer join are
exceptions because moving such projections above join produces the same results and such move is
automatically done by SQL optimizer. Also, if a calculation cannot guarantee that NULL will not be returned in
the result, then intermediate result will not be materialized.

Below is an example where coalesce was used to guarantee that order_key will not return NULL which triggers
materialization.

Table 20:
Calculation that will not trigger materialization
SELECT * FROM customer C LEFT OUTER
JOIN (
SELECT L.order_key FROM orders O
JOIN lineitem L ON O.order_key =
L.order_key
) ON C.cust_key = L.order_key;

Calculation that will trigger materialization

SELECT * FROM customer C LEFT OUTER
JOIN (
SELECT coalesce(L.order_key, 0)
FROM orders O JOIN lineitem L ON
O.order_key = L.order_key
) ON C.cust_key = L.order_key;

A possible workaround of avoiding materialization of intermediate result is by adding a generated column for
the constants or calculated values.

SAP HANA Troubleshooting and Performance Analysis Guide

106 PUBLIC Root Causes And Solutions
Table 21:
Problematic query
SELECT * FROM customer C LEFT OUTER
JOIN (
SELECT 1 const FROM orders O
JOIN lineitem L ON O.order_key =
L.order_key ) ON C.cust_key = L.const

Workaround
ALTER TABLE ORDERS ADD (const INTEGER
GENERATED ALWAYS AS 1);
SELECT * FROM customer C LEFT OUTER
JOIN (
SELECT * FROM orders O JOIN
lineitem L ON O.order_key =
L.order_key ) ON C.cust_key = L.const;

3.9.12.3 Avoid Inefficient Predicates in EXISTS/IN

This section will list inefficient predicate conditions.

3.9.12.3.1 Disjunctive EXISTS Predicate

If possible avoid disjunctive EXISTS predicates.

When an EXISTS or NOT EXISTS predicate is connected with other predicates through OR, it is internally
mapped to left outer join. Since left outer join processing is more expensive than inner join processing in
general, we recommend avoiding such disjunctive EXISTS predicates if possible. Also, avoid using OR to
connect EXISTS or NOT EXISTS predicates with other predicates if possible.

Below is an example of avoiding disjunctive EXISTS predicate.

Table 22:
Problematic query
SELECT * FROM T WHERE EXISTS (SELECT
* FROM S WHERE S.a = T.a AND S.b = 1)
OR EXISTS (SELECT * FROM S WHERE S.a
= T.a AND S.b = 2);

Workaround
SELECT * FROM T WHERE EXISTS (SELECT
* FROM S WHERE S.a = T.a AND (S.b = 1
OR S.b = 2));

This tuning tip also applies to the following cases:

● Filter predicate inside NOT EXISTS predicate accessing multiple tables

● Filter predicate inside disjunctive EXISTS predicate accessing multiple tables

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 107
3.9.12.3.2 NOT IN Predicate

The NOT IN predicate is much more expensive to process than NOT EXISTS, it is recommended to use NOT
EXISTS instead of NOT IN if possible.

In general, for NOT IN, entire subquery is processed first before the processing the query as a whole, matching
up based on the condition provided. However, with NOT EXISTS, true or false will be returned while checking
for provided condition. Unless subquery results is very small, using NOT EXISTS is much faster than NOT IN.
( Same applies for EXISTS/IN ).

Below is an example of avoiding the NOT IN predicate. Please note that the transformation in the example is
not valid in general. It is valid only if there is no null values in the columns of interest. The transformation is
automatically applied by SQL optimizer if all columns of interest have NOT NULL constraints declared
explicitly.

Table 23:
NOT IN query
SELECT * FROM T WHERE a NOT IN
(SELECT a FROM S);

Possibly equivalent query in some cases

SELECT * FROM T WHERE NOT EXISTS
(SELECT * FROM S WHERE S.a = T.a);

3.9.12.4 Avoid Set Operations

UNION ALL, UNION, INTERSECT and EXCEPT are not natively supported by the column engine, avoiding them
may improve performance.

Examples of avoiding UNION, INTERSECT and EXCEPT are shown below. Note that the transformations in the
examples are not valid in general. They are valid only if there is no null value in the columns of interest. The
transformations for INTERSECT and EXCEPT are automatically applied by SQL optimizer if all columns of
interest have NOT NULL constraints declared explicitly.

Table 24:
UNION query
SELECT a, b FROM T UNION SELECT a, b
FROM S;

Possibly equivalent query in some cases

SELECT DISTINCT COALESCE(T.a, S.a) a,
COALESCE(T.b, S.b) b FROM T FULL
OUTER JOIN S ON T.a = S.a AND T.b =
S.b;

INTERSECT query
SELECT a, b FROM T INTERSECT SELECT
a, b FROM S;

SAP HANA Troubleshooting and Performance Analysis Guide

108 PUBLIC Root Causes And Solutions
 Possibly equivalent query in some cases
SELECT DISTINCT T.a a, T.b b FROM T
JOIN S ON T.a = S.a AND T.b = S.b;

EXCEPT query
SELECT a, b FROM T EXCEPT SELECT a, b
FROM S;

Possibly equivalent query in some cases

SELECT DISTINCT T.a a, T.b b FROM T
WHERE NOT EXISTS (SELECT * FROM S
WHERE T.a = S.a AND T.b = S.b);

3.9.12.5 Improve Performance For Multiple Column Join

Creating a concatenated column index can improve query performance when multiple columns are involved in
a join.

One way to optimize this type of query is to create concatenated column index explicitly. However, please note
that creation of the index will increase memory consumption.

Below is an example of query that needs concatenated columns and syntax to create the needed
concatenated columns.

Table 25:
Problematic Query
SELECT M.year, M.month,
SUM(T.ship_amount)
FROM T JOIN M ON T.year = M.year AND
T.month = M.month
GROUP BY M.year, M.month;

Workaround
CREATE INDEX T_year_month ON T(year,
month);
CREATE INDEX M_year_month ON M(year,
month);

The following hints can be tried to optimize performance for multiple column join predicate:

WITH HINT(OPTIMIZE_METAMODEL) --prefer creation of concatenated attribute during

query processing time
WITH HINT(NO_OPTIMIZE_METAMODEL) --avoid creation of concatenated attribute

3.9.12.6 Using Hints to Alter a Query Plan

This section lists hints that may be used to alter a query plan.

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 109
3.9.12.6.1 Changing an Execution Engine Decision

You can use hints to change which engine is used to execute your query.

The SQL optimizer chooses a column search execution engine (Join Engine or Olap Engine) based on the cost-
model. For various reasons, optimal plan may not be executed using the best engine. The following hints can
be applied to change the execution engine used.

WITH HINT(USE_OLAP_PLAN) -- to prefer OLAP engine (if possible)

WITH HINT(NO_USE_OLAP_PLAN) -- to prefer JOIN engine

3.9.12.6.2 Changing Query Transformation

Hints can be used to apply or remove query transformation rules.

By examining the plan, you can apply different hints to alter query transformation as needed. By using
NO_<hint>, then can be disabled to produce opposite effect. For a full list of available hints, please refer to
HINT Details in the SAP HANA SQL and System Views Reference.

Table 26: Hint: GROUPING_SIMPLIFICATION

Before After

Table 27: Hint: GROUPING_REMOVAL

Before After

SAP HANA Troubleshooting and Performance Analysis Guide

110 PUBLIC Root Causes And Solutions
Related Information

SAP HANA SQL and System Views Reference

3.9.12.6.3 Changing Operator Order

Hints can be used to change the order of operators during plan generation.

By examining the plan, you can apply different hints to change the operator order as needed. By using
NO_<hint>, they can be disabled to produce opposite effect. For a full list of available hints, please refer to
HINT Details in the SAP HANA SQL and System Views Reference.

Table 28: Hint: AGGR_THRU_JOIN

Before After

Table 29: Hint: PREAGGR_BEFORE_JOIN

Before After

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 111
Table 30: Hint: FILTER_THRU_JOIN

Before After

Related Information

SAP HANA SQL and System Views Reference

3.9.12.6.4 Choosing Preferred Algorithm

Hints can be used to select preferred algorithms for execution ( column engine vs row engine ).

By using NO_<hint>, they can be disabled to produce opposite effect. For a full list of available hints, please
refer to HINT Details in the SAP HANA SQL and System Views Reference.

Table 31: Hint: CS_JOIN

Before After

SAP HANA Troubleshooting and Performance Analysis Guide

112 PUBLIC Root Causes And Solutions
Table 32: Hint: CS_ORDERBY

Before After

Table 33: Hint: RANGE_JOIN

Before After

Related Information

SAP HANA SQL and System Views Reference

3.9.12.6.5 Using Alter System Command To Pin Hints To

Problematic Query

A hint table can be used to persist the binding between a query and hints.

If you want to persist the binding between a query and hints or you are unable to append hints to a query
during runtime (that is application generated queries ), a hint table can be used. For a full list of available hints,
please refer to HINT Details in the SAP HANA SQL and System Views Reference.

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 113
Related Information

SAP HANA SQL and System Views Reference

3.9.13 Creation of Indexes on Non-Primary Key Columns

Create indexes on non-primary key columns to enhance the performance of some queries using the index
adviser.

SAP HANA automatically creates indexes for all primary key columns.

Indexing the primary key columns is usually sufficient because queries typically put filter conditions on
primary key columns. When filter conditions are on non-key fields and tables have many records, creating an
index on the non-primary key columns may improve the performance.

Tip
To check whether there is an index for a column, in the Plan Visualizer, in the properties of a column, see the
Inverted Index entry. Alternatively, you can also see the system view M_INDEXES.

You can create indexes on non-primary key columns to enhance the performance of some queries, particularly
highly selective queries on non-primary key columns.

Use the index adviser to find out for which tables and columns indexing would be most valuable. The
indexAdvisor.py script is part of a SAP HANA system installation and runs from the command line. It is located
in the $DIR_INSTANCE/exe/python_support directory.

There is a trade-off between indexing and memory consumption: While indexing non-primary key columns can
make query execution faster, the downside is that memory consumption increases. The index adviser takes
this trade-off into account: In dynamic mode, the index adviser looks for the tables and columns that are used
most often. The higher the selectivity is, that is, the more different values are in the column, the higher are the
performance gains from indexing the columns.

To create indexes on non-primary columns, use the following SQL statement:

CREATE INDEX <name> ON <table> (<column>)

Related Information

SAP HANA SQL and System Views Reference

SAP HANA Troubleshooting and Performance Analysis Guide

114 PUBLIC Root Causes And Solutions
3.9.14 Create Analytic Views

Analytic views are used to model data that includes measures. For example, transactional fact table
representing sales order history would include measures for quantity, price, and so on.

Procedure

1. Launch SAP HANA studio.

2. In SAP HANA System view, expand the content node.
3. In the navigator, select a package where you want to create the new calculation view.

4. In the context menu of the package, select New Analytic View .

5. Provide name and description.
6. Choose Finish.

Modeler launches a new analytiv view editor with the semantics node, star join node and the data
foundation.
7. Add data sources.
a. Select the star join node.
b. In the context menu, choose Add Objects.
c. In Find Data Sources dialog, enter the name of the data source.

Note
You can only dimensions (attribute views) as a data source in star join node of analytic views.

d. Select the required attribute view from the list.

e. Choose OK.
8. Define the central fact table in the data foundation node.

Continue modeling the analytic view with a cube structure, which includes attributes and measures. The
input to the star join node must provide the central fact table.

a. Select the data foundation node.

b. In the context menu, choose Add Objects.
c. In Find Data Sources dialog, enter the name of the data source and select it from the list.

Note
You cannot add column views to the Data Foundation.

d. Choose OK.

Note
You can to add the same table again in Data Foundation using table aliases in the editor.

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 115
 e. If there are more than one table in the data foundation node, specify the central table (fact table) from
which the modeler must derive the measures. Selec the Data foundation node and define the property,
Central Entity in the Properties pane.
9. Define output columns.
a. Select the data foundation node or star join node.
a. In the Details pane, select the columns that you want to add to the output of the node.
b. In the context menu, choose Add To Output.

Note
If you want to add all columns from the data source to the output, in the context menu of the data
source, choose Add All To Output.

10. In the Star Join node, create joins to join the attribute views with the fact table (star schema).

You can also create a temporal joins between date fields of the fact table to an interval (to and from) field
of the attribute view.

Restriction
Self-joins are not supported. While creating joins, ensure that a table does not appear twice in any join
path. A join path is the set of joins that links the fact table to other tables.

While creating joins between analytic view and attribute view:

○ The same table cannot be used in the join path of analytic view and attribute view
○ The table of the attribute view which is linked to the fact table should not have an alias table

11. Define attributes and measures

a. Select the Semantics node.
b. Choose the Columns tab.
c. In the Local section, select a output column.
d. In the Type dropdown list, select a measure or attribute.
e. If you want to hide the measure of attribute in the reporting tool, select the Hidden checkbox.

Note
The Shared tab page shows attributes from the attribute views that are used in the analytic view.

12. Activate the analytic view:

○ If you are in the SAP HANA Modeler perspective:
○ Save and Activate - to activate the current view and redeploy the affected objects if an active
version of the affected object exists. Otherwise only current view gets activated.
○ Save and Activate All - to activate the current view along with the required and affected objects.

Note
You can also activate the current view by selecting the view in the SAP HANA Systems view and
choosing Activate in the context menu. The activation triggers validation check for both the client
side and the server side rules. If the object does not meet any validation check, the object activation
fails.

SAP HANA Troubleshooting and Performance Analysis Guide

116 PUBLIC Root Causes And Solutions
 ○ If you are in the SAP HANA Development perspective:
1. In the Project Explorer view, select the required object.
2. In the context menu, select Team Activate .

Note
The activation triggers the validation check only for the server side rules. Hence, if there are
any errors on the client side, they are skipped and the object activation goes through if no error
found at the server side.

Note
If an active version of the affected objects exist, activating the current view redeploys the affected
objects. In the SAP HANA Modeler perspective, even if the affected object redeployment fails, the
current view activation might go through. However, in the SAP HANA Development perspective, if any
of the affected objects redeployment fails, the current view activation also fails.

Note
While modeling an analytic view, if you have also opened and edited an attribute view that is used in the
analytic view, then close and reopen the analytic view editor to see any changes that you have made to
the attribute view. For more information, see SAP Note 1783668 .

13. Assign Changes

a. In the Select Change dialog, either create a new ID or select an existing change ID that you want to use
to assign your changes.
b. Choose Finish.
For more information on assigning changes, see chapter SAP HANA Change Recording of the SAP
HANA Developer Guide.
14. Choose Finish.

Next Steps

After creating an analytic view, you can perform certain additional tasks to obtain the desired output. The table
below lists the additional tasks that you can perform to enrich the analytic view.

Table 34: Working With View Nodes

Requirement Task to Perform

If you want to filter the output of the data foundation node. Filter Output of Data Foundation
Node.

Table 35: Working With Attributes and Measures

Requirement Task to perform

If you want to count the count the number of distinct values for a set of attribute col­ Create Counters
umns.

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 117
 Requirement Task to perform

If you want to create new output columns and calculate its values at runtime using Create Calculated Columns
an expression.

If you want to restrict measure values based on attribute restrictions. Create Restricted Columns

If you want to assign semantic types to provide more meaning to attributes and Assign Semantics
measures in analytic views.

If you want to parameterize attribute views and execute them based on the values Create Input Parameters
users provide at query runtime.

If you want to, for example, filter the results based on the values that users provide Assign Variables
to attributes at runtime.

If you want associate measures with currency codes and perform currency conver­ Associate Measures with Currency
sions.

If you want associate measures with unit of measures and perform unit conversions. Associate Measures with Unit of
Measure

If you want to group related measures together in a folder. Group Related Measures.

Table 36: Working With Ananlytic View Properties

Requirement Task to perform

If you want to filter the view data either using a fixed client value or using a session Filter Data for Specific Clients
client set for the user.

If you want to execute time travel queries on analytic views. Enable Information Views for Time
Travel Queries

If you want to invalidate or remove data from the cache after specific time intervals. Invalidate Cached Content

If you want to maintain object label texts in different languages. Maintain Modeler Objects in Multi­
ple Languages

If you do not recommend using an anlytic view. Deprecate Information Views

3.9.15 Developing Procedures

SQL in SAP HANA includes extensions for creating procedures, which enables you to embed data-intensive
application logic into the database, where it can be optimized for performance (since there are no large data
transfers to the application and features such as parallel execution is possible). Procedures are used when
other modeling objects, such as calculation views, are not sufficient; procedures are also often used to support
the database services of applications that need to write data into the database.

Some of the reasons to use procedures instead of standard SQL:

● SQL is not designed for complex calculations, such as for financials.

● SQL does not provide for imperative logic.
● Complex SQL statements can be hard to understand and maintain.
● SQL queries return one result set. Procedures can return multiple result sets.
● Procedures can have local variables, eliminating the need to explicitly create temporary tables for
intermediate results.

SAP HANA Troubleshooting and Performance Analysis Guide

118 PUBLIC Root Causes And Solutions
Procedures can be written in the following languages:

● SQLScript: The language that SAP HANA provides for writing procedures.
● R: An open-source programming language for statistical computing and graphics, which can be installed
and integrated with SAP HANA.

There are additional libraries of procedures, called Business Function Library and Predictive Analysis Library,
that can be called via SQL or from within another procedure.

Tools for Developing Procedures

The following describes the tools integrated with SAP HANA studio that allow you to evaluate, revise, and
optimize stored procedures:

SQLScript Editor

The SQLScript editor studio allows you to create, edit, and debug stored procedures. Within the SQLScript
editor you can use the Semantic Code Completion feature. The semantic code completion feature is a context
based search tool that lists suggested catalog object and local variables that assist you with developing
accurate stored procedures in a faster and more efficient matter. You can quickly identify valid objects
reducing errors during activation. Code completion proposals take into consideration SQLScript grammar,
context specific schemas, and textual input.

The suggested objects are derived from the following origins:

● Catalog Objects: such as schemas, views, table functions, procedures, scalar functions, synonyms
● Local Variables: such as input and output parameters, declared scalar variables
● Database Artifacts

Suggested objects are listed alphabetically, according to the following format:

[icon] artifact_name - artifact_type (artifact_context), for example DUMMY - Table

(<Schema name>)

The list of proposals contain syntactic and semantic proposals listed in the following order:

1. Local Variables
2. Catalog Objects (maximum of 50 suggestions)
3. Keywords

Note
Objects selected from the proposed list might be automatically inserted as quoted identifier based on the
SQLScript language guidelines For example, if the object contains special characters, or lower and upper
case characters.

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 119
Debugger

In addition to the creating and editing procedures, the SQLScript editor includes debugging capabilities.

You can also use the Navigator view in the Modeler perspective to build procedures, but there are no
debugging capabilities. You should only use this method:

● If you need to develop a procedure using a local table type as an input or output parameter. A local table
type is created within the SAP HANA Systems procedure tool and for only the current procedure. If you
can use a global table type, then use the SQLScript Editor.
● If you need to edit a procedure previously created in the Navigator view that contains table type
parameters.

SQL Extensions for Procedures

SQL includes the following statements for enabling procedures:

● CREATE TYPE: Creates a table types, which are used to define parameters for a procedure that represent
tabular results. For example:

CREATE TYPE tt_publishers AS TABLE (

publisher INTEGER,
name VARCHAR(50),
price DECIMAL,
cnt INTEGER);

● CREATE PROCEDURE: Creates a procedure. The LANGUAGE clause specifies the language you are using to
code the procedure. For example:

CREATE PROCEDURE ProcWithResultView(IN id INT, OUT o1 CUSTOMER)

LANGUAGE SQLSCRIPT READS SQL DATA WITH RESULT VIEW ProcView AS
BEGIN
o1 = SELECT * FROM CUSTOMER WHERE CUST_ID = :id;
END;

● CALL: Calls a procedure. For example:

CALL getOutput (1000, 'EUR', NULL, NULL);

Related Information

SAP HANA SQL and System Views Reference

SAP HANA SQLScript Reference
SAP HANA R Integration Guide
SAP HANA Business Function Library (BFL) Reference
SAP HANA Predictive Analysis Library (PAL) Reference

SAP HANA Troubleshooting and Performance Analysis Guide

120 PUBLIC Root Causes And Solutions
3.9.16 Application Function Library (AFL)

You can dramatically increase performance by executing complex computations in the database instead of at
the application sever level.

SAP HANA provides several techniques to move application logic into the database, and one of the most
important is the use of application functions. Application functions are like database procedures written in C+
+ and called from outside to perform data intensive and complex operations. Functions for a particular topic
are grouped into an application function library (AFL), such as the Predictive Analytical Library (PAL) or the
Business Function Library (BFL).

Currently, all AFLs are delivered in one archive (that is, one SAR file with the name AFL<version_string>.SAR).

Note
The AFL archive is not part of the SAP HANA appliance, and must be installed separately by an
administrator. For more information about installing the AFL archive, see the SAP HANA Server Installation
and Update Guide.

Security Considerations

● User and Schema

During startup, the system creates the user _SYS_AFL, whose default schema _SYS_AFL.

Note
The user and its schema _SYS_AFL are created during a new installation or update process if they do
not already exist.

All AFL objects, such as areas, packages, functions, and procedures, are created under this user and
schema. Therefore, all these objects have fully specified names in the form of _SYS_AFL.<object name>.
● Roles
For each AFL library, there is a role. You must be assigned this role to execute the functions in the library.
The role for each library is named: AFL__SYS_AFL_<AREA NAME>_EXECUTE. For example, the role for
executing PAL functions is AFL__SYS_AFL_AFLPAL_EXECUTE.

Note
There are 2 underscores between AFL and SYS.

Note
Once a role is created, it cannot be dropped. In other words, even when an area with all its objects is
dropped and recreated during system start-up, the user still keeps the role that was previously granted.

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 121
Related Information

SAP HANA Server Installation and Update Guide

3.9.17 Aspects of Scalability

Before you decide how to scale your SAP HANA implementation, there are a number of aspects that need to
be considered, such as scaling data, performance, applications, and hardware.

Scaling the Data

One technique you can use to deal with planned data growth is to purchase more physical RAM than is initially
required, to set the allocation limit according to your needs, and then to increase it over time to adapt to your
data. Once you have reached the physical limits of a single server, you can scale out over multiple machines to
create a distributed SAP HANA system. You can do this by distributing different schemas and tables to
different servers (complete data and user separation). However, this is not always possible, for example, when
a single fact table is larger than the server's RAM size.

The most important strategy for scaling your data is data partitioning. Partitioning supports the creation of
very large tables (billions of rows) by breaking them into smaller chunks that can be placed on different
machines. Partitioning is transparent for most SQL queries and other data manipulations.

For more information, see the section on managing tables.

Scaling Performance

SAP HANA's performance is derived from its efficient, parallelized approach. The more computation cores
your SAP HANA server has, the better overall system performance.

Scaling performance requires a more detailed understanding of your workload and performance expectations.
Using simulations and estimations of your typical query workloads, you can determine the expected load that
a typical SAP HANA installation may comfortably manage. At the workload level, a rough prediction of
scalability can be established by measuring the average CPU utilization while the workload is running. For
example, an average CPU utilization of 45% may indicate that the system can be loaded 2X before showing a
significant reduction in individual query response time.

For more information, see the sections on workload management and performance analysis.

Scaling the Application

Partitioning can be used to scale the application as it supports an increasing number of concurrent sessions
and complex analytical queries by spreading the calculations across multiple hosts. Particular care must be

SAP HANA Troubleshooting and Performance Analysis Guide

122 PUBLIC Root Causes And Solutions
taken in distributing the data so that the majority of queries match partitioning pruning rules. This
accomplishes two goals: directing different users to different hosts (load balancing) and avoiding the network
overhead related to frequent data joins across hosts.

Scaling Hardware

SAP HANA is offered in a number of ways – in the form of an on-premise appliance, delivered in a number of
different configurations and "sizes" by certified hardware partners or by using the tailored data center
integration model, and as part of a cloud-based service. This creates different system design options with
respect to scale-up and scale-out variations. To maximize performance and throughput, SAP recommends
that you scale up as far as possible (acquire the configuration with the highest processor and memory
specification for the application workload), before scaling out (for deployments with even greater data volume
requirements).

Note
The SAP HANA hardware partners have different building blocks for their scale-out implementations.
Therefore, you should always consult with your hardware partner when planning your scale-out strategy.

3.9.18 Further Recommendations

In addition to the general recommendations for improving SAP HANA database performance, for specific
scenarios, you can use further features and best practices to improve performance.

If appropriate, you can take the following actions to improve performance:

● For any required long-running transactions, you can use the SQL command ALTER SYSTEM RECLAIM
VERSION SPACE to trigger the row-store garbage collector to free up memory space and enhance system
responsiveness.
● For multicolumn join scenarios, use dynamic joins rather than standard joins.
In a dynamic join, the elements of a join condition between two data sources are defined dynamically
based on the fields requested by the client query. It is used to improve the performance by reducing the
number of records to be processed by the join node.
● When inserting or loading a large number of rows into a table that has a TEXT or SHORTTEXT column or
uses a FULLTEXT INDEX, merge the delta of the table for better search performance.
● When loading data from CSV files using the IMPORT FROM command, use THREADS and BATCH to enable
parallel loading and commit many records at once. In general, for column tables, a good setting to use is
10 parallel loading threads, with a commit frequency of 10,000 records or greater. You can also use TABLE
LOCK, which locks the entire table and bypasses the delta table. Tables locks are only recommended for
initial loads.

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 123
3.10 Application Performance Analysis

This section provides a high-level guide for analyzing the performance of SAP HANA-based applications.

When you have a SAP HANA-based application with unsatisfying performance you should start a systematic
investigation considering the various layers that are involved. Here we focus on the database layer;
approaches for analyzing UI and application server performance are described in the respective documents.

3.10.1 SQL Trace Analysis

The first step in application performance analysis is to figure out if the database layer is causing performance
problems for your application at all.

Context

You should analyze how many and which database calls are made and what their contribution to the overall
application performance is. This should be done within the context of a given user interface step or
transaction.

Procedure

1. Start the tracing of database calls.

2. Run the application from its user interface or with any other driver.

Both, SAP HANA studio and SAP HANA Web-based Development Workbench provide two main tracing
tools, namely SQL trace and Expensive statements trace.

A convenient way to narrow the trace analysis to the scope of a user interface step or transaction is to use
the passport-based filtering of Expensive statements trace in the SAP HANA Web-based Development
Workbench, which also offers aggregated statistics to quickly answer above questions.

Deep tracing (including complete execution plans) is provided by Plan Trace in SAP HANA studio.
3. Terminate the tracing and review aggregated and individual results.
4. As a result of this investigation you might see some indicators for bad application logic creating excessive
load on the database such as:
○ Too many database calls (per transaction/UI step)
○ Many identical executions, for example repeated identical selects
○ Too many records returned (per execution or in total)
○ Too many columns or all columns of a row selected
○ Inefficient statement reuse, that is, statements that need to be optimized over and over again
○ One or more database calls with unexpected bad performance, so you should further investigate
those calls

SAP HANA Troubleshooting and Performance Analysis Guide

124 PUBLIC Root Causes And Solutions
Related Information

Plan Trace [page 170]

Analyzing SQL Traces [page 151]
Analysis Tools in SAP HANA Web-based Developer Workbench [page 183]

3.10.2 Statement Measurement

Once you have determined which SQL statements are problematic you should first perform a sound
measurement in order to get reliable performance numbers and to make sure that indeed your statement(s)
are causing the issues and not the current state of your SAP HANA system.

Procedure

1. Execute your statement(s) and measure their performance (in particular response time).

Both, SAP HANA studio and SAP HANA Web-based Development Workbench offer basic measurement of
SQL statements. In addition, the SAP HANA Web-based Development Workbench implicitly supports
observing the system state and also executing repeated measurements. In SAP HANA studio you have to
use the administration perspective in order to examine the system state.
2. Check your SAP HANA system status for disturbing conditions, such as high load, high resource usage
and so on
3. In case of disturbing conditions repeat your measurement from step 1.
4. Repeat your measurements until you get stable results without major variations (for example, 3 stable
executions in a row).

Note
Higher initial response times could be an indicator of caches that are not properly filled. Depending on
your business needs you can decide whether this is acceptable or not.

5. Once you have a stable result you may also acquire a detailed SAP HANA engine trace which will allow for
a deeper analysis.

Results

As a result of this activity you have reliable data for your query performance, both for initial query execution
performance (possibly cold execution) and stabilized execution performance (warm execution).

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 125
Related Information

Statement Performance Analysis [page 87]

3.10.3 Data Analysis

The easiest way to analyze a poorly performing SQL statement is to look at the data flow as this can be
matched with the understanding of the business/application needs.

Procedure

1. Check the result size, that is the number of records and number of columns, returned by your SQL
statement and compare it with the actual business needs.

As a result of this investigation you might restrict the result size by changing the application logic (for
example, the number columns selected in the field list or by applying additional filters in the WHERE
clause) with regards to its database call.
2. Check the usage of underlying tables, meaning the set of tables used, their size and the number of entries
selected from those tables, and compare it with your understanding of the business needs.
As a result of this investigation you might identify tables that should not be involved at all and adapt your
statement or the underlying database logic accordingly (for example, by checking joins and join types).
SAP HANA studio offers data flow analysis with the Tables Used view in the SAP HANA Plan Visualizer
perspective.
3. Check the data flow of your statement, that is, the order in which tables are joined, how filters are applied
and the size of intermediate results.

As a result of this investigation you may identify:

○ inefficient join orders (starting with table A and not with table B, when that is the much smaller result
set)
○ unexpectedly missing filters (that is, intermediate selections which seem too broad).

You can then adapt your statement or underlying database logic as necessary.

SAP HANA studio offers data flow analysis with the Graphical Plan view in the SAP HANA Plan Visualizer
perspective.

Related Information

Tables Used [page 168]

Analyzing SQL Execution with the Plan Visualizer [page 158]

SAP HANA Troubleshooting and Performance Analysis Guide

126 PUBLIC Root Causes And Solutions
3.10.4 Source Analysis

Before conducting a detailed technical analysis it is recommended to first analyze source artifacts (models,
scripts, views) for possible performance-impacting aspects.

Context

The actual activities depend on the respective artifact (models, scripts, views).

Procedure

1. Graphical Calculation Models

SAP HANA studio offers dedicated performance debugging capabilities as part of the Modeler for
graphical calculation models; furthermore the Plan Visualizer perspective provides deeper inspection of
these models.
a. Check the effect of the attributes, parameters and filters used in the query on the execution time.
b. Check that the information is combined and aggregated in the correct way as required by the business
scenario, for example:
○ Is the aggregation behavior consistent with the intended semantics?
○ Is the join cardinality correct for each join?
○ Are dynamic joins used?
○ Is the result always aggregated?
○ Has Join Optimization been enabled?
○ Do analytical privileges influence performance?
c. Check whether any performance-relevant execution hints are activated, for example:
○ Enforce execution via a specified database engine
○ Enforce upper bound for parallelization
d. Check whether modeling constructs are used that are known to be costly, for example:
○ Calculated join attributes
○ Complicated filter expressions
○ Mixed models that involve engine switches
e. Check whether intermediate steps produce reasonable results (in terms of size).
2. SQL Script

The Statement Statistics view (part of the Plan Visualizer) support analysis of SQLScript.
a. Check if your procedure fits with the given guidelines
For example, see Developing Procedures
b. Analyze your script for most expensive steps / statements
3. Plain SQL
a. Check if your statement fits with the guidelines for SQL Query Design
For example, see SQL Query Design

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 127
Results

As a result of this activity you either figured out and solved problems at the source level or you now have to
continue with an even deeper technical analysis.

Related Information

Developing Procedures [page 118]

Technical Analysis [page 128]
Statement Statistics [page 162]

3.10.5 Technical Analysis

The deepest level of performance analysis addresses the technical details of a database statement execution.

Context

You should follow this in order to track down problematic performance symptoms from which you can derive
possible root causes at higher levels.

There are a number of tools that can support you:

● SAP HANA studio offers the Plan Visualizer perspective which allows for deep technical analysis.
Dedicated views and filters support the analysis along numerous dimensions.
● Furthermore, there are multiple tools allowing for even deeper analysis such as Explain Plan and the
numerous tracing tools.

Procedure

1. Inspect aggregated execution KPIs (execution time, resource consumption, distribution characteristics) in
order to figure out in which aspect or dimension you might look for possible issues.
2. For any suspicious KPI, track down the KPI to the deepest possible level manifesting the symptom.
3. From there, try to correlate the symptom with the cause, in higher level statement elements.
4. You might restrict the scope of your analysis by focusing on the critical path, or on a specific time interval,
system node, engine, or execution plan operator.

SAP HANA Troubleshooting and Performance Analysis Guide

128 PUBLIC Root Causes And Solutions
Related Information

Analyzing SQL Execution with the Plan Visualizer [page 158]

Analyzing SQL Execution with the Plan Explanation [page 157]

3.11 System Hanging Situations

This section explains what checks to carry out and how to perform those checks if your SAP HANA instance is
hanging. The database is said to be hanging when it no longer responds to queries that are executed against it.

Context

The source of the system standstill might be related to any of the components involved, for example, the
storage, OS and hardware, network, SAP HANA database or the application layer. For troubleshooting it is
essential to collect information about the context of the active threads in the SAP HANA database.

As SQL statements cannot usually be executed for analysis, you should perform the following steps if it is still
possible to log on to the OS of the master host (for example, as the <sid>adm user). Also see SAP Note
1999020 on "SAP HANA: Troubleshooting when database is no longer reachable" for further specific steps and
guidance on pro-active or reactive actions you can take.

Procedure

1. Collect a runtime dump (see SAP Note 1813020 - How to generate a runtime dump on SAP HANA)
2. Collect CPU, memory and other information about threads currently running in the system by executing
the command top -H and taking a screenshot of the output.

Note
Transparent Huge Pages (THP) is only applicable to servers with Intel processors.

3. Please provide this information when logging an incident with SAP support and avoid restarting the
database as otherwise retrospective analysis might not always be possible.

Related Information

SAP Note 1999020

SAP Note 1813020
Transparent Huge Pages [page 130]

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 129
CPU Power Saving [page 131]

3.11.1 Transparent Huge Pages

Transparent Huge Pages (THP) is a feature that is generally activated for the Linux kernel shipped with SUSE
Linux Enterprise Server (SLES) 11 and Red Hat Enterprise Linux (RHEL) 6 versions certified for usage with SAP
HANA. Due to the special manner of SAP HANA's memory management, using THP may lead to hanging
situations and performance degradations.

Context

You experience that your SAP HANA database does not react anymore, that the CPU load is high and/or
severe performance issues. Additionally, Transparent Huge Pages are activated on your SAP HANA servers.
Until further notice, SAP strongly recommends that you disable Transparent Huge Pages on all your SAP
HANA servers.

Note
Transparent Huge Pages is supported on Intel-based hardware platforms only.

See the SAP Notes below for further reference:

● SAP Note 2131662 - Transparent Huge Pages (THP) on SAP HANA Servers
● SAP Note 1824819 SAP HANA DB: Recommended OS settings for SLES 11 / SLES for SAP Applications 11
SP2
● SAP Note 1954788 SAP HANA DB: Recommended OS settings for SLES 11 / SLES for SAP Applications 11
SP3
● SAP Note 2013638 SAP HANA DB: Recommended OS settings for RHEL 6.5
● SAP Note 2136965 SAP HANA DB: Recommended OS settings for RHEL 6.6

Note
the following checks and steps should be performed on all host of the affected SAP HANA system. They
have to be executed as the root user in the Linux shell.

Procedure

1. To check whether Transparent Huge Pages are activated and currently being used by processes, execute
the below commands in the Linux shell:
a. cat /sys/kernel/mm/transparent_hugepage/enabled
b. cat /proc/meminfo | grep AnonHugePages

SAP HANA Troubleshooting and Performance Analysis Guide

130 PUBLIC Root Causes And Solutions
 If the first output shows "[always] madvise never" that is "always" is shown inside square brackets, then
THPs are activated according to the value of the relevant Linux Kernel parameter. If the second output
shows a value greater than 0 Kb (Kilobytes), then this is evidence that some processes use THPs.
2. If both steps show that Transparent Huge Pages are activated or in use, deactivate them by executing the
commands below in the Linux shell:
a. echo never > /sys/kernel/mm/transparent_hugepage/enabled
b. cat /sys/kernel/mm/transparent_hugepage/enabled
The first command will deactivate Transparent Huge Pages by setting the relevant Kernel parameter to
"never". Right after executing this command the hanging situation will be resolved. You can verify whether
the Kernel parameter has the correct value by executing the second command which should show the
output "always madvise [never]" where the new value is in square brackets.
3. Note that unless the value of the Kernel parameter has not been persisted, it is only set/valid until the host
is restarted the next time. To prevent toggling the value from [never] back to [always] causing THPs to be
activated again, use one of the below strategies to persist it:
a. Add the parameter "transparent_hugepage=never" to the kernel boot line in the /etc/grub.conf file
of the Bootloader.
b. Integrate the parameter "transparent_hugepage=never" within your system boot scripts (for
example,. /etc/rc.local).

Related Information

SAP Note 2131662

SAP Note 1824819
SAP Note 1954788
SAP Note 2013638
SAP Note 2136965

3.11.2 CPU Power Saving

The Linux Kernel shipped with SUSE Linux Enterprise Server (SLES) 11 and Red Hat Enterprise Linux (RHEL) 6
versions certified for usage with SAP HANA contain a new cpuidle driver for recent Intel CPUs. This driver
leads to a different behavior in C-states switching and causes performance degradations.

Context

See the SAP Notes below for further reference:

● SAP Note 1824819 SAP HANA DB: Recommended OS settings for SLES 11 / SLES for SAP Applications 11
SP2
● SAP Note 1954788 SAP HANA DB: Recommended OS settings for SLES 11 / SLES for SAP Applications 11
SP3

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 131
● SAP Note 2013638 SAP HANA DB: Recommended OS settings for RHEL 6.5
● SAP Note 2136965 SAP HANA DB: Recommended OS settings for RHEL 6.6

Procedure

1. Check if the recommended driver is enabled and whether the CPU power safe mode is activated. Execute
the following command as root user in the Linux shell.

cat /sys/devices/system/cpu/cpuidle/current_driver

The correct value for the cpuidle driver should be "acpi_idle". If so no further steps are required. In case
the output shows the wrong value "intel_idle" please follow the steps in SAP Notes.
2. Check the CPU power save mode by running the following command.

cat /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor

The correct value for the CPU power save mode should be "performance". If the output shows at lease one
line with "ondemand" please follow the steps in SAP Note 1890444 - Slow HANA system due to CPU power
save mode.

Related Information

SAP Note 1824819

SAP Note 1954788
SAP Note 2013638
SAP Note 2136965
SAP Note 1890444

3.12 Troubleshoot System Replication

If system replication appears to slow down transaction processing you can check the network and disk I/O on
the secondary site.

Often due to network related issues or a disk I/O related issue on the secondary site, system replication can
slow down transaction processing in the following cases:

● Asynchronous replication is configured over long distances.

● Multitier system replication is configured and a tier 3 system is attached.
● SYNC/SYNCMEM replication mode is configured over short distances

The following troubleshooting steps can help you determine and resolve the underlying cause.

SAP HANA Troubleshooting and Performance Analysis Guide

132 PUBLIC Root Causes And Solutions
Check If Log Can Be Shipped In Time

● Check if log shipping is significantly slower than local log write (SYNC/SYNCMEM)
● Check Async Buffer Full Count (ASYNC)
● These checks can be used to verify that the problem is really related to system replication
● For an overview of basic system replication KPIs see HANA_Replication_SystemReplication_Overview.txt
(from SAP Note 1969700). It is based on the system view M_SERVICE_REPLICATION and can be used to
compare log shipping time to local log write time.
● The SAP HANA Cockpit can also be used to check system replication KPIs. For more information see, Tile
Catalog: SAP HANA System Replication in the SAP HANA Administration Guide.

Check If Data Load Can Be Handled By Network Link

● Estimate the amount of data/log shipped from the primary site

● Compare this with the available bandwidth (the recommended bandwidth is 10 Gbit/s)
● Do a network performance test. For example, use the open source IPERF tool or similar to measure the
real application network performance
● To estimate the required bandwidth for Data/Log shipping, use
HANA_Replication_SystemReplication_Bandwidth.txt (from SAP Note 1069700), which is based on the I/O
statistics from the primary site. We recommend executing this SQL statement when system replication is
disabled.

Check Network Configuration (long distance)

Increasing the TCP window size can result in better network utilization and higher throughput.

● Collect network information on bandwidth and latency

● If the bandwidth can handle load check if the network is shared and whether other applications may be
interfering with performance.
● Check the network utilization profile for the network link to see if the maximum capacity of the network
has been reached
● If the network is not fully utilized check the linux kernel TCP configuration with sysctl –a | egrep
“net.core|net.ipv4.tcp”
● Check that windows scaling is set to the default value of 1. net.ipv4.tcp_window_scaling = 1
● Check whether the max size can be increased for net.ipv4.tcp_wmem and net.ipv4.tcp_rmem
● The BDP tells you what TCP window size is needed to use the network link fully.
Calculate the Bandwidth Delay Product (BDP): Bandwidth * Latency (for example, BDP = 50ms * 3 Gbps =
19.2 MB).

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 133
Check Disk I/O on Secondary Site

Slow disk I/O on the secondary can postpone releasing log buffers on primary, which results in wait situations
on the primary. You can:

● Use a Disk Performance Test Tool

○ Execute fsperf on log volume
○ For example:
$ fsperf /usr/sap/TST/SYS/global/hdb/log/mnt00001/hdb00002
● Check the Monitoring view
○ If SQL is not available, use command line tools (this has to be done for each individual service)
○ For example: $ hdbcons "statreg print -n M_VOLUME_IO_TOTAL_STATISTICS -h"
○ A runtime dump also contains I/O statistics, which you can see with: $ hdbcons “runtimedump
dump”

Caution
Technical expertise is required to use hdbcons. To avoid incorrect usage, use hdbcons only with
the guidance of SAP HANA development support.

● Check I/O relevant tables in the proxy schema of the corresponding secondary site, which provide SQL
access on the primary site on statistic views of the secondary. For more information, see Monitoring
Secondary Sites in the SAP HANA Administration Guide.

Related Information

SAP Note 1969700

3.13 Network Performance and Connectivity Problems

This section covers the troubleshooting of problems on the network layer.

In cases where a subjectively slow performing system behaviour is experienced, but a first analysis of the SAP
HANA resource utilization does not reveal any obvious culprits, it is often necessary to analyze the network
performance between the SAP HANA server host(s) and SAP Application Server(s) / Non-ABAP clients, SAP
HANA nodes (inter-node communication in SAP HANA scale-out environments), or, in an SAP HANA system
replication scenario, between primary and secondary site.

3.13.1 Network Performance Analysis on Transactional Level

The following section should help you to perform an in-depth investigation on the network performance of
specific clients.

SAP HANA Troubleshooting and Performance Analysis Guide

134 PUBLIC Root Causes And Solutions
Prerequisites

SYSTEM Administrator access using SAP HANA studio or hdbsql.

Procedure

1. Use the monitoring view M_SQL_CLIENT_NETWORK_IO to analyse figures about client and server elapsed
time as well as message sizes for client network messages.

Sample Code

SELECT * FROM M_SQL_CLIENT_NETWORK_IO

In case a long execution runtime is observed on the application server side and the corresponding
connections on the SAP HANA side do not show expensive operations, an overview of the total processing
time spent on client side and SAP HANA server side can retrieved by executing the above SQL query. By
default, collection of statistics related to client network I/O is regulated by the following parameter
sql_client_network_io in the indexserver.ini file, which must be set to on (true):

Please note that this parameter change implies a certain performance overhead and should only be active
for the duration of the troubleshooting activity.

An example result of the above mentioned query is shown here:

CLIENT_DURATION and SERVER_DURATION contain the values in microseconds. The difference between
CLIENT_DURATION and SERVER_DURATION makes the total transfer time of the result
(SEND_MESSAGE_SIZE in bytes). This allows you to see whether the transfer time from the SAP HANA
server to the client host is exceptionally high.
2. Run SQL: “HANA_Network_Clients” from the SQL Statement collection attached to SAP Note 1969700

Another important KPI is the Round Trip Time (RTT) from server to client. In order to analyze this figure
the SQL statement "HANA_Network_Clients" from the collection attached to SAP Note 1969700 can be

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 135
 used. As this SQL statement is using the view M_EXPENSIVE_STATEMENTS, the expensive statements
trace needs to be active in SAP HANA studio Administration editor Trace Configuration tab:

Once the trace is activated and the long running statements is re-executed, the information to be
extracted from the M_EXPENSIVE_STATEMENTS view is STATEMENT_HASH:

With the Statement Hash '7c4a13b071f030f1c0d178ab9cf82c37' (please note that this one is only an
example statement hash) of the SQL statement to be analyzed, the SQL “HANA_Network_Clients” can be
modified in a way that a corresponding Statement Hash is used to restrict the selection:

Sample Code

…
SELECT /* Modification section */
TO_TIMESTAMP('1000/01/01 18:00:00', 'YYYY/MM/DD HH24:MI:SS')
BEGIN_TIME,
TO_TIMESTAMP('9999/12/31 18:10:00', 'YYYY/MM/DD HH24:MI:SS')
END_TIME,
'%' HOST,
'%' PORT,
'%' SERVICE_NAME,
'%' CLIENT_HOST,
'7c4a13b071f030f1c0d178ab9cf82c37' STATEMENT_HASH,
.
.
.*/
FROM
DUMMY

Which provides the following result:

The KPI AVG_RTT_MS is of importance and should not show values significantly higher than ~ 1,5 ms.
3. For further options, please refer to SAP KBA 2081065.

SAP HANA Troubleshooting and Performance Analysis Guide

136 PUBLIC Root Causes And Solutions
Related Information

SAP KBA 2081065

SAP Note 1969700

3.13.2 Stress Test with NIPING

The SAP NIPING tool is a powerful tool which can be used to perform specific network stability tests.

Prerequisites

You must have OS level access to SAP HANA host and client host.

Procedure

Read SAP Note 500235 - Network Diagnosis with NIPING.

A stress test with SAP's NIPING tool may be performed in order to confirm the high network latency (or
bandwidth exhaustion).

Related Information

SAP Note 500235

3.13.3 Application and Database Connectivity Analysis

There are a number of ways to identify possible root-causes for network communication issues between your
application and the SAP HANA instance it is connecting to.

Prerequisites

You have access to both the application and SAP HANA instance.

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 137
Procedure

1. On an ABAP application server, check the following:

a. Run transaction OS01 - Database - Ping x10

If a connection to the database cannot be established over a longer period of time by an SAP ABAP
application work process, the work process is terminated. First, the work process enters the
reconnect state in which it constantly tries to connect to the database, after a predefined amount of
retries fail, the work process terminates. In this case the connectivity from the SAP application server
to the SAP HANA server must be verified.
b. Run transaction SE38 - Report ADBC_TEST_CONNECTION

If a specific database connection is failing, the report ADBC_TEST_CONNECTION offers a connectivity

check for each defined database connection
c. Check for common errors in SAP KBA 2213725

If an application is facing communication issues with the SAP HANA server, on client side the
connectivity issue may be indicated by several 10709-errors, mostly short dumps. The error 10709 is
generic but the error text of the short-dumps contains the error information that was returned by the
server. Some root causes may be found in unfavorable parameter settings, either on client or on
server side, some may be caused by a faulty network.

In most cases, a short-dump with characteristics is raised:

Sample Code

Category Installation Errors

Runtime Errors DBSQL_SQL_ERROR
Except. CX_SY_OPEN_SQL_DB

The "Database error text" gives you a first hint as to what might have caused the issue. For an
overview of the most common errors in this context and detailed explanations of how to resolve them,
see SAP KBA 2213725.
2. On non-ABAP applications, check the following SAP Notes:
a. SAP Note 1577128 - Supported clients for SAP HANA

In case issues occur on non-ABAP client connections to a remote SAP HANA instance, it is of
importance to make sure that a supported client is used.
b. SAP Note 1993254 - Collecting ODBC Trace

A typical SAP HANA ODBC connection failure is indicated by an error with the following prefix:

Sample Code

[LIBODBCHDB SO][HDBODBC]....

The most common errors are documented in SAP Notes and KBAs. Please enter the error texts in
xSearch to find a solution for a specific error. If no SAP Notes can be found which outline the root
cause of a specific error message you can record an ODBC trace to gain more insight.
a. SAP KBA 2081065 - Troubleshooting SAP HANA Network

SAP HANA Troubleshooting and Performance Analysis Guide

138 PUBLIC Root Causes And Solutions
 In case the error occurs sporadically, it is useful to perform a long-term stress test between the client
and SAP HANA server to confirm the network's stability. For more information, see SAP Note 500235.
To examine the exact traffic from the TCP/IP layer, a tcpdump can be recorded which shows what
packets are sent and received and which packets were rejected or required a retransmission.
3. Generic Smart Data Access troubleshooting steps are:
a. To verify that the issue might not be a flaw in the SAP HANA studio, always try to connect via the 'isql'
tool on the SAP HANA host directly as the SAP HANA<sid>adm.
b. Make sure all libraries can be accessed on OS level by the SAP HANA <sid>adm user (PATH
environment variable)
c. In the SAP HANA <sid>adm home directory (cd $home) check that the correct host and port, and
username and password combinations are used in the .odbc.ini file.
d. Check that the LD_LIBRARY_PATH environment variable of the SAP HANA <sid>adm user contains
the unixODBC and remote DB driver libraries.

For example, with a Teradata setup, the LD_LIBRARY_PATH should contain the following
paths: .../usr/local/unixODBC/lib/:/opt/teradata/client/15.00/odbc_64/lib...

Connections from the SAP HANA server to remote sources are established using the ODBC interface
(unixODBC). For more information, see the SAP HANA Administration Guide.
4. Microsoft SQL Server 2012 - Specific Smart Data Access Troubleshooting
a. Make sure unixODBC 2.3.0 is used. A higher version leads to a failed installation of the Microsoft
ODBC Driver 11 for SQL Server 2012.
b. The version can be verified with the following command executed as the SAP HANA <sid>adm user
on the SAP HANA host: 'isql --version'
c. Check SAP KBA 2233376 - HANA Troubleshooting Tree - Smart Data Access
5. Teradata - Specific Smart Data Access Troubleshooting
a. Check SAP KBA 2078138 - "Data source name not found, and no default driver specified" when
accessing a Teradata Remote Source through HANA Smart Data Access

Under certain circumstances it is necessary to adjust the order of the paths maintained in
LD_LIBRARY_PATH

Related Information

SAP KBA 2213725

SAP Note 1577128
SAP Note 1993254
SAP KBA 2081065
SAP Note 500235
SAP KBA 2233376
SAP KBA 2078138
SAP xSearch
SAP HANA Administration Guide

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 139
3.13.4 SAP HANA System Replication Communication
Problems

Problems during initial setup of the system replication can be caused by incorrect configuration, incorrect
hostname resolution or wrong definition of the network to be used for the communication between the
replication sites.

Context

System replication environments depend on network bandwidth and stability. In case communication
problems occur between the replication sites (for example, between SITE A and SITE B), the first indication of
a faulty system replication setup will arise.

Procedure

1. Check the nameserver tracefiles.

The starting point for the troubleshooting activity are the nameserver tracefiles. The most common errors
found are:

Sample Code

e TNS TNSClient.cpp(00800) : sendRequest dr_secondaryactivestatus to

<hostname>:<system_replication_port> failed with NetException.
data=(S)host=<hostname>|service=<service_name>|(I)drsender=2|
e sr_nameserver TNSClient.cpp(06787) : error when sending request
'dr_secondaryactivestatus' to <hostname>:<system_replication_port>:
connection broken,location=<hostname>:<system_replication_port>
e TrexNetBuffer BufferedIO.cpp(01151) : erroneous channel ### from ##### to
<hostname>:<system_replication_port>: read from channel failed; resetting
buffer

Further errors received from the remote side:

Sample Code

Generic stream error: getsockopt, Event=EPOLLERR - , rc=104: Connection

reset by peer
Generic stream error: getsockopt, Event=EPOLLERR - , rc=110: Connection
timed out

It is important to understand that if those errors suddenly occur in a working system replication
environment, they are often indicators of problems on the network layer. From an SAP HANA perspective,
there is nothing that could be toggled, as it requires further analysis by a network expert. The
investigation, in this case, needs to focus on the TCP traffic by recording a tcpdump in order to get a rough
understanding how TCP retransmissions, out-of-order packets or lost packets are contributing to the
overall network traffic. How a tcpdump is recorded is described in SAP Note 1227116. As these errors are

SAP HANA Troubleshooting and Performance Analysis Guide

140 PUBLIC Root Causes And Solutions
 not generated by the SAP HANA server, please consider consulting your in-house network experts or your
hardware vendor before engaging with SAP Product Support.
2. Set the parameter sr_dataaccess to debug.

In the Administration editor of the SAP HANA studio open the Configuration tab. In the [trace] section of
the indexserver.ini file set the parameter sr_dataaccess = debug. This parameter enables a more
detailed trace of the components involved in the system replication mechanisms.

Related Information

SAP Note 1227116 - Creating network traces

3.13.5 SAP HANA Inter-Node Communication Problems

This section contains analysis steps that can be performed to resolve SAP HANA inter-node communication
issues.

Procedure

1. If communication issues occur between different nodes within an SAP HANA scale-out environment,
usually the SAP HANA tracefiles contain corresponding errors.

A typical error recorded would be:

Sample Code

e TrexNet Channel.cpp(00343) : ERROR: reading from channel ####

<IP_of_remote_host:3xx03> failed with timeout error; timeout=60000 ms
elapsed
e TrexNetBuffer BufferedIO.cpp(01092) : channel #### from : read from
channel failed; resetting buffer

To understand those errors it is necessary to understand which communications parties are affected by
this issue. The <IP:port> information from the above mentioned error already contains valuable
information. The following shows the port-numbers and the corresponding services which are listening on
those ports:

Sample Code

Nameserver 3<instance_no>01
Preprocessor 3<instance_no>02
Indexserver 3<instance_no>03
Webdispatcher 3<instance_no>06
XS Engine 3<instance_no>07
Compileserver 3<instance_no>10

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 141
 Interpreting the above error-message it is safe to assume that the affected service was failing to
communicate with the indexserver of one node in the scale-out system. Please note that these errors are
usually caused by network problems and should be analyzed by the person responsible for OS or network
or the network team of the hardware vendor.

2. In SAP HANA studio open Administration Performance Threads and check the column Thread
Status for Network Poll, Network Read, Network Write

In case the Threads tab in the SAP HANA Studio Administration editor shows many threads with the state
Network Poll, Network Read or Network Write, this is a first indication that the communication (Network
I/O) between the SAP HANA services or nodes is not performing well and a more detailed analysis of the
possible root causes is necessary. For more information about SAP HANA Threads see SAP KBA 2114710.
3. Run SQL: "HANA_Network_Statistics"

As of SAP HANA SPS 10 the view M_HOST_NETWORK_STATISTICS provides SAP HANA host related
network figures. The SAP HANA SQL statement collection from SAP Note 1969700 contains SQL:
“HANA_Network_Statistics” which can be used to analyze the network traffic between all nodes within a
SAP HANA scale-out system.

Sample Code

----------------------------------------------------------------------------
---------
|HOST |SEG_RECEIVED |BAD_SEG_RCV|BAD_SEG_PCT|SEG_SENT |SEG_RETRANS|
RETRANS_PCT|
----------------------------------------------------------------------------
---------
|hostnam| 163965201| 282| 0.00017| 340922924|
19520| 0.00572|
----------------------------------------------------------------------------
---------

For a detailed documentation of the figures from this output please refer to the documentation section of
the SQL statement SQL: “HANA_Network_Statistics”
4. Run SAP HANA Configuration Mini Checks from SAP KBA 1999993.

The "Network" section of the mini-check results contains the following checks:

Sample Code

----------------------------------------------------------------------------
------------------
|CHID |DESCRIPTION |HOST |VALUE |
EXPECTED_VALUE|C|SAP_NOTE|
----------------------------------------------------------------------------
------------------
|**** |NETWORK | |
| | | |
| | | |
| | | |
| 1510|Avg. intra node send throughput (MB/s) |hostnam|444 |>=
120 | | 2222200|
| 1512|Avg. inter node send throughput (MB/s) |hostnam|never |>=
80 | | 2222200|
| 1520|Retransmitted TCP segments (%) | |0.00571 |<=
0.10000 | | 2222200|
| 1522|Bad TCP segments (%) | |0.00017 |<=
0.01000 | | 2222200|

SAP HANA Troubleshooting and Performance Analysis Guide

142 PUBLIC Root Causes And Solutions
 ----------------------------------------------------------------------------
------------------

The results usually contain an "expected value" (which provides a certain "rule of thumb" value) and a
"value" field which represents the actual value recorded on the system. If the recorded value is breaching
the limitations defined by the expected value, the "C" column should be flagged with an 'X'. You can then
check the note for this item referenced in the column SAP_NOTE.

Related Information

SAP KBA 2114710 - FAQ: SAP HANA Threads and Thread Samples
SAP Note 1969700
SAP KBA 1999993

SAP HANA Troubleshooting and Performance Analysis Guide

Root Causes And Solutions PUBLIC 143
4 Tools and Tracing

This section gives you an overview of the available tools and tracing options that are available.

4.1 System Performance Analysis

As a first step to resolving SAP HANA performance issues, you can analyze detailed aspects of system
performance in the SAP HANA studio on the Performance tab of the Administration editor.

When analyzing system performance issues, the information provided on the Performance tab enables you to
focus your analysis on the following questions:

● What and how many threads are running, what are they working on, and are any of these threads blocked?
● Are any sessions blocking current transactions?
● Are any operations running for a significantly long time and consuming a lot of resources? If so, when will
they be finished?
● How do different hosts compare in terms of performance?

On the Performance tab, you can take certain actions to improve performance, including canceling the
operations that cause blocking situations.

4.1.1 Thread Monitoring

You can monitor all running threads in your system in the Administration editor on the Performance
Threads sub-tab. It may be useful to see, for example, how long a thread is running, or if a thread is blocked
for an inexplicable length of time.

Thread Display

By default, the Threads sub-tab shows you a list of all currently active threads with the Group and sort filter
applied. This arranges the information as follows:

● Threads with the same connection ID are grouped.

● Within each group, the call hierarchy is depicted (first the caller, then the callee).
● Groups are displayed in order of descending duration.

On big systems with a large number of threads, this arrangement provides you with a more meaningful and
clear structure for analysis. To revert to an unstructured view, deselect the Group and sort checkbox or change
the layout in some other way (for example, sort by a column).

SAP HANA Troubleshooting and Performance Analysis Guide

144 PUBLIC Tools and Tracing
Thread Information

Detailed information available on the Threads sub-tab includes the following:

● The context in which a thread is used

This is indicated by the thread type. Important thread types are SqlExecutor and PlanExecutor.
SqlExecutor threads handle session requests such as statement compilation, statement execution, or
result fetching issued by applications on top of SAP HANA. PlanExecutor threads are used to process
column-store statements and have an SqlExecutor thread as their parent.

Note
With revision 56, PlanExecutor threads were replaced by JobWorker threads.

Note
The information in the Thread Type column is only useful to SAP Support for detailed analysis.

● What a thread is currently working on

The information in Thread Detail, Thread Method, and Thread Status columns is helpful for analyzing what
a thread is currently working on. In the case of SqlExecutor threads, for example, the SQL statement
currently being processed is displayed. In the case of PlanExecutor threads (or JobWorker threads as of
revision 56), details about the execution plan currently being processed are displayed.

Note
The information in the Thread Detail, Thread Method, and Thread Status columns is only useful to SAP
Support for detailed analysis.

● Information about transactionally blocked threads

A transactionally blocked thread is indicated by a warning icon ( ) in the Status column. You can see
detailed information about the blocking situation by hovering the cursor over this icon.
A transactionally blocked thread cannot be processed because it needs to acquire a transactional lock
that is currently held by another transaction. Transactional locks may be held on records or tables.
Transactions can also be blocked waiting for other resources such as network or disk (database or
metadata locks).
The type of lock held by the blocking thread (record, table, or metadata) is indicated in the Transactional
Lock Type column.
The lock mode determines the level of access other transactions have to the locked record, table, or
database. The lock mode is indicated in the Transactional Lock Type column.
Exclusive row-level locks prevent concurrent write operations on the same record. They are acquired
implicitly by update and delete operations or explicitly with the SELECT FOR UPDATE statement.
Table-level locks prevent operations on the content of a table from interfering with changes to the table
definition (such as drop table, alter table). DML operations on the table content require an intentional
exclusive lock, while changes to the table definition (DDL operations) require an exclusive table lock.
There is also a LOCK TABLE statement for explicitly locking a table. Intentional exclusive locks can be
acquired if no other transaction holds an exclusive lock for the same object. Exclusive locks require that no
other transaction holds a lock for the same object (neither intentional exclusive nor exclusive).
For more detailed analysis of blocked threads, information about low-level locks is available in the columns
Lock Wait Name, Lock Wait Component and Thread ID of Low-Level Lock Owner. Low-level locks are locks
acquired at the thread level. They manage code-level access to a range of resources (for example, internal

SAP HANA Troubleshooting and Performance Analysis Guide

Tools and Tracing PUBLIC 145
 data structures, network, disk). Lock wait components group low-level locks by engine component or
resource.
The Blocked Transactions sub-tab provides you with a filtered view of transactionally blocked threads.

Monitoring and Analysis Features

To support monitoring and analysis, you can perform the following actions on the Threads sub-tab:

● See the full details of a thread by right-clicking the thread and choosing Show Details.
● End the operations associated with a thread by right-clicking the thread and choosing Cancel Operations.

Note
This option is not available for threads of external transactions, that is those with a connection ID of -1.

● Jump to the following related objects by right-clicking the thread and choosing Navigate To <related
object> :
○ Threads called by and calling the selected thread
○ Sessions with the same connection ID as the selected thread
○ Blocked transactions with the same connection ID as the selected thread
● View the call stack for a specific thread by selecting the Create call stacks checkbox, refreshing the page,
and then selecting the thread in question.

Note
The information contained in call stacks is only useful to SAP Support for detailed analysis.

● Activate the expensive statements trace, SQL trace, or performance trace by choosing Configure Trace
<required trace> .
The Trace Configuration dialog opens with information from the selected thread automatically entered
(application and user).

Note
If the SQL trace or expensive statements trace is already running, the new settings overwrite the
existing ones. If the performance trace is already running, you must stop it before you can start a new
one.

Related Information

SQL Trace [page 151]

Performance Trace [page 180]
Expensive Statements Trace [page 154]
SAP HANA SQL and System Views Reference

SAP HANA Troubleshooting and Performance Analysis Guide

146 PUBLIC Tools and Tracing
4.1.2 Blocked Transaction Monitoring

Blocked transactions, or transactionally blocked threads, can impact application responsiveness. They are
indicated in the Administration editor on the Performance Threads tab. You can see another
representation of the information about blocked and blocking transactions on the Blocked Transactions sub-
tab.

Information About Blocked Transactions

Blocked transactions are transactions that are unable to be processed further because they need to acquire
transactional locks (record or table locks) that are currently held by another transaction. Transactions can
also be blocked waiting for other resources such as network or disk (database or metadata locks).

The type of lock held by the blocking transaction (record, table, or metadata) is indicated in the Transactional
Lock Type column.

The lock mode determines the level of access other transactions have to the locked record, table, or database.
The lock mode is indicated in the Transactional Lock Type column.

Exclusive row-level locks prevent concurrent write operations on the same record. They are acquired implicitly
by update and delete operations or explicitly with the SELECT FOR UPDATE statement.

Table-level locks prevent operations on the content of a table from interfering with changes to the table
definition (such as drop table, alter table). DML operations on the table content require an intentional
exclusive lock, while changes to the table definition (DDL operations) require an exclusive table lock. There is
also a LOCK TABLE statement for explicitly locking a table. Intentional exclusive locks can be acquired if no
other transaction holds an exclusive lock for the same object. Exclusive locks require that no other transaction
holds a lock for the same object (neither intentional exclusive nor exclusive).

For more detailed analysis of blocked transactions, information about low-level locks is available in the
columns Lock Wait Name, Lock Wait Component and Thread ID of Low-Level Lock Owner. Low-level locks are
locks acquired at the thread level. They manage code-level access to a range of resources (for example,
internal data structures, network, disk). Lock wait components group low-level locks by engine component or
resource.

Monitoring and Analysis Features

To support monitoring and analysis, you can perform the following actions on the Blocked Transactions sub-
tab:

● Jump to threads and sessions with the same connection ID as a blocked/blocking transaction by right-
clicking the transaction and choosing Navigate To <related object> .
● Activate the performance trace, SQL trace, or expensive statements trace for the blocking transaction
(that is the lock holder) by choosing Configure Trace <required trace> .
The Trace Configuration dialog opens with information from the selected thread automatically entered
(application and user).

SAP HANA Troubleshooting and Performance Analysis Guide

Tools and Tracing PUBLIC 147
 Note
If the SQL trace or expensive statements trace is already running, the new settings overwrite the
existing ones. If the performance trace is already running, you must stop it before you can start a new
one.

Related Information

SQL Trace [page 151]

Performance Trace [page 180]
Expensive Statements Trace [page 154]
SAP HANA SQL and System Views Reference

4.1.3 Session Monitoring

You can monitor all sessions in your landscape in the Administration editor on the Performance
Sessions sub-tab.

Session Information

The Sessions sub-tab allows you to monitor all sessions in the current landscape. You can see the following
information:

● Active/inactive sessions and their relation to applications

● Whether a session is blocked and if so which session is blocking
● The number of transactions that are blocked by a blocking session
● Statistics like average query runtime and the number of DML and DDL statements in a session
● The operator currently being processed by an active session (Current Operator column).

Note
In earlier revisions, you can get this information from the SYS.M_CONNECTIONS monitoring view with
the following statement:

SELECT CURRENT_OPERATOR_NAME FROM M_CONNECTIONS WHERE CONNECTION_STATUS =

'RUNNING'

Tip
To investigate sessions with the connection status RUNNING, you can analyze the SQL statements being
processed in the session. To see the statements, ensure that the Last Executed Statement and Current
Statement columns are visible. You can then copy the statement into the SQL console and analyze it using

SAP HANA Troubleshooting and Performance Analysis Guide

148 PUBLIC Tools and Tracing
 the Explain Plan and Visualize Plan features. It is also possible to use the SQL plan cache to understand and
analyze SQL processing.

Monitoring and Analysis Features

To support monitoring and analysis, you can perform the following actions on the Sessions sub-tab:

● Cancel a session by right-clicking the session and choosing Cancel Session...

● Jump to the following related objects by right-clicking the session and choosing Navigate To <related
object> :
○ Threads with the same connection ID as the selected session
○ Blocked transactions with the same connection ID as the selected session
● Activate the performance trace, SQL trace, or expensive statements trace by choosing Configure Trace
<required trace> .
The Trace Configuration dialog opens with information from the selected session automatically entered
(application and user).

Note
If the SQL trace or expensive statements trace is already running, the new settings overwrite the
existing ones. If the performance trace is already running, you must stop it before you can start a new
one.

Related Information

SQL Trace [page 151]

Performance Trace [page 180]
Expensive Statements Trace [page 154]
SAP HANA SQL and System Views Reference

4.1.4 Job Progress Monitoring

Certain operations in SAP HANA typically run for a long time and may consume a considerable amount of
resources. You can monitor long-running jobs in the Administration editor on the Performance Job
Progress sub-tab.

By monitoring the progress of long-running operations, for example, delta merge operations and data
compression, you can determine whether or not they are responsible for current high load, see how far along
they are, and when they will finish.

The following information is available, for example:

SAP HANA Troubleshooting and Performance Analysis Guide

Tools and Tracing PUBLIC 149
● Connection that triggered the operation (CONNECTION_ID)
● Start time of the operation (START_TIME)
● Steps of the operation that have already finished (CURRENT_PROGRESS)
● Maximum number of steps in the operation (MAX_PROGRESS)

For more information about the operations that appear on the Job Progress sub-tab, see system view
M_JOB_PROGRESS.

Related Information

SAP HANA SQL and System Views Reference

4.1.5 Load Monitoring

A graphical display of a range of system performance indicators is available in the Administration editor on the
Performance Load sub-tab.

You can use the load graph for performance monitoring and analysis. For example, you can use it to get a
general idea about how many blocked transactions exist now and in the past, or troubleshoot the root cause of
slow statement performance.

Related Information

SAP HANA Troubleshooting and Performance Analysis Guide

4.2 SQL Statement Analysis

A key step in identifying the source of poor performance is understanding how much time SAP HANA spends
on query execution. By analyzing SQL statements and calculating their response times, you can better
understand how the statements affect application and system performance.

You can analyze the response time of SQL statements with the following traces:

● SQL trace
From the trace file, you can analyze the response time of SQL statements.
● Expensive statements trace
On the Performance Expensive Statements Trace tab, you can view a list of all SQL statements that
exceed a specified response time.

In addition to these traces, you can analyze the SQL plan cache, which provides a statistical overview of what
statements are executed in the system.

SAP HANA Troubleshooting and Performance Analysis Guide

150 PUBLIC Tools and Tracing
4.2.1 Analyzing SQL Traces

The SQL trace allows you to analyze the response time of SQL statements within an object.

Procedure

1. In the Administration editor, choose the Trace Configuration trace and edit the SQL trace.
2. In the Trace Configuration dialog box, specify a name for the trace file, set the trace status to Active, and
specify the required trace and user filters.
3. Choose Finish.
4. Run the application or SQL statements you want to trace.
5. Re-open the SQL trace configuration and set the trace status to Inactive.
6. Choose Finish.
7. Choose the Diagnosis Files tab and open the trace file you created.
8. Choose Show Entire File.
9. Analyze the response time of the relevant SQL statements to identify which statements negatively affect
performance.
The SQL statements in the trace file are listed in order of execution time. To calculate the response time of
a specific SQL statement, calculate the difference between the times given for # tracing
PrepareStatement_execute call and # tracing finished PrepareStatement_execute.

4.2.1.1 SQL Trace

The SQL trace collects information about all SQL statements executed on the index server and saves it in a
trace file for further analysis. It is inactive by default.

Information collected by the SQL trace includes overall execution time of each statement, the number of
records affected, potential errors (for example, unique constraint violations) that were reported, the database
connection being used, and so on. So the SQL trace is a good starting point for understanding executed
statements and their potential effect on the overall application and system performance, as well as for
identifying potential performance bottlenecks at statement level.

Note
In the system database of a multiple-container system, there is no index server. Here, the name server
functions as the index server. This means that the SQL trace in the system database runs on the name
server.

SQL trace information is saved as an executable python program that you can access on the Diagnosis Files
tab of the Administration editor.

SAP HANA Troubleshooting and Performance Analysis Guide

Tools and Tracing PUBLIC 151
 Note
Writing SQL trace files can impact database performance significantly. They also consume storage space
on the disk. Therefore, it is not recommended that you leave the SQL trace enabled all the time.

You activate and configure the SQL trace in the Administration editor on the Trace Configuration tab.

Related Information

SQL Trace Options [page 152]

4.2.1.2 SQL Trace Options

Several options are available for configuring the SQL trace.

Trace Levels

You can configure the following trace levels:

Table 37:
Trace Level Description

NORMAL All statements that have finished successfully are traced with detailed information such as
executed timestamp, thread ID, connection ID, and statement ID.

ERROR All statements that returned errors are traced with detailed information such as executed
timestamp, thread ID, connection ID, and statement ID.

ERROR_ROLLBACK All statements that are rolled back are traced with detailed information such as executed
timestamp, thread ID, connection ID and statement ID.

ALL All statements including status of normal, error, and rollback are traced with detailed infor­
mation such as executed timestamp, thread ID, connection ID and statement ID.

ALL_WITH_RESULTS In addition to the trace generated with trace level ALL, the result returned by select state­
ments is also included in the trace file.

Note
An SQL trace that includes results can quickly become very large.

Caution
An SQL trace that includes results may expose security-relevant data, for example,
query result sets.

SAP HANA Troubleshooting and Performance Analysis Guide

152 PUBLIC Tools and Tracing
Other Configuration Options

The following additional configuration options are available:

Table 38:
Option Description

Trace file User-specific name for the trace file

If you do not enter a user-specific file name, the file name is generated according to the fol­
lowing default pattern:

DB_<dbname>/sqltrace_$HOST_${PORT}_${COUNT:3}.py, where:
● DB_<dbname> is the sub-directory where the trace file is written if you are running on a
tenant database
● $HOST is the host name of the service (for example, indexserver)
● $PORT is the port number of the service
● $COUNT:3 is an automatically generated 3-digit number starting with 000 that incre­
ments by 1 and serves as a file counter when several files are created.

User, application, object, and Filters to restrict traced statements to those of particular database or application users and
statement filters
applications, as well as to certain statement types and tables.

All statements matching the filter criteria are recorded and saved to the specified trace file.

Flush limit During tracing, the messages of a connection are buffered. As soon as the flush limit num­
ber of messages is buffered (or if the connection is closed), those messages are written to
the trace file.

4.2.2 Analyzing Expensive Statements Traces

The expensive statements trace allows you to identify which SQL statements require a significant amount of
time and resources.

Procedure

1. In the Administration editor, choose the Trace Configuration trace and edit the expensive statements
trace.
2. In the Trace Configuration dialog box, set the trace status to Active and specify a threshold execution time
in microseconds.
The system will identify any statements that exceed this threshold as expensive statements.
3. Choose Finish.
4. Run the application or SQL statements you want to trace.

5. Choose the Performance Expensive Statements Trace tab.

SAP HANA Troubleshooting and Performance Analysis Guide

Tools and Tracing PUBLIC 153
6. Analyze the displayed information to identify which statements negatively affected performance.
For each SQL statement, the following columns are especially useful for determining the statement's
impact on performance:
○ START_TIME
○ DURATION_MICROSEC
○ OBJECT_NAME (names of the objects accessed)
○ STATEMENT_STRING
○ CPU_TIME

4.2.2.1 Expensive Statements Trace

Expensive statements are individual SQL statements whose execution time exceeded a configured threshold.
The expensive statements trace records information about these statements for further analysis. It is inactive
by default.

You activate and configure the expensive statements trace in the Administration editor on either the Trace
Configuration tab or the Performance tab. Information about recorded expensive statements is displayed on
the Performance tab.

If in addition to activating the expensive statements trace, you enable per-statement memory tracking, the
expensive statements trace will also show the peak memory size used to execute the expensive statements.

Note
You can also monitor expensive statements and configure the expensive statements trace in the SAP HANA
cockpit.

Related Information

Expensive Statements Trace Options [page 154]

Setting a Memory Limit for SQL Statements [page 32]

4.2.2.2 Expensive Statements Trace Options

Several options are available for configuring the expensive statements trace.

Table 39:
Option Description

Threshold duration Threshold execution time in microseconds (default 1,000,000)

User, application filters, and Filters to restrict traced statements to those of particular database or application users, as
object filters well as to certain applications and tables

SAP HANA Troubleshooting and Performance Analysis Guide

154 PUBLIC Tools and Tracing
 Option Description

Passport trace level If you are activating the expensive statements trace as part of an end-to-end trace scenario
with the Process Monitoring Infrastructure (PMI), you can specify the passport trace level
as an additional filter.

This means that only requests that are marked with a passport of the specified level are
traced. For more information see, SAP Library for NetWeaver on SAP Help Portal.

Note
Process tracing is possible only for components in the ABAP and Business Objects
stacks.

Trace parameter values In SQL statements, field values may be specified as parameters (using a "?" in the syntax).
If these parameter values are not required, you can deselect the Trace parameter values
checkbox to reduce the amount of data traced.

Additional Parameters

You can configure the expensive statement trace further using the following properties in the
expensive_statement section of global.ini configuration file:

Table 40:
Property Description

trace_flush_interval Number of records after which trace file is flushed

threshold_cpu_time Threshold CPU time of statement execution in microsec­

onds

Note
Resource tracking and CPU time tracking must also be
enabled. You can do this by configuring the correspond­
ing parameters in the resource_tracking section
of the global.ini file.

threshold_memory Threshold memory usage of statement execution in bytes

Note
Resource tracking and memory tracking must also be en­
abled. You can do this by configuring the corresponding
parameters in the resource_tracking section of
the global.ini file.

Related Information

Process Monitoring with PMI (Process Monitoring Infrastructure)

SAP HANA Troubleshooting and Performance Analysis Guide

Tools and Tracing PUBLIC 155
4.2.3 Analyzing SQL Execution with the SQL Plan Cache

The SQL plan cache is a valuable tool for understanding and analyzing SQL processing.

Before it is executed, every SQL statement is compiled to a plan. Once a plan has been compiled, it is better to
reuse it the next time the same statement is executed, instead of compiling a new plan every time. The SAP
HANA database provides an object, the SQL plan cache, that stores plans generated from previous executions.
Whenever the execution of a statement is requested, a SQL procedure checks the SQL plan cache to see if
there is a plan already compiled. If a match is found, the cached plan is reused. If not, the statement is
compiled and the generated plan is cached.

As the SQL plan cache collects statistics on the preparation and execution of SQL statements, it is an
important tool for understanding and analyzing SQL processing. For example, it can help you to find slow
queries, as well as measure the overall performance of your system.

You can access the SQL plan cache in the Administration editor on the Performance tab. The two system views
associated with the SQL plan cache are M_SQL_PLAN_CACHE_OVERVIEW and M_SQL_PLAN_CACHE.

The SQL plan cache contains a lot of information. Filtering according to the following columns can help you
identify statements that are more likely to be causing problems and/or could be optimized:

Table 41:
Column Description

TOTAL_EXECUTION_TIME The total time spent for all executions of a plan

This helps to identify which statements are dominant in terms of time.

AVG_EXECUTION_TIME The average time it takes to execute a plan execution

This can help you identify long-running SQL statements.

EXECUTION_COUNT The number of times a plan has been executed

This can help you identify SQL statements that are executed more frequently than ex­
pected.

TO­ The total number of waiting locks

TAL_LOCK_WAIT_COUNT
This can help you identify SQL statements with high lock contention.

USER_NAME The name of the user who prepared the plan and therefore where the SQL originated
(ABAP/index server/statistics server)

For a full list of all SQL cache columns including descriptions, see the documentation for the system views
M_SQL_PLAN_CACHE_OVERVIEW and M_SQL_PLAN_CACHE in the SAP HANA SQL and System Views
Reference.

Related Information

SAP HANA SQL and System Views Reference

SAP HANA Troubleshooting and Performance Analysis Guide

156 PUBLIC Tools and Tracing
4.3 Query Plan Analysis

In SAP HANA, to identify queries that are inefficiently processed, you can both technically and logically analyze
the steps SAP HANA took to process those queries.

From a technical perspective, analyzing query plans allows you to identify long running steps, understand how
much data is processed by the operators, and see whether data is processed in parallel. However, if you
understand the idea and purpose behind the query, you can also analyze query plans from a logical
perspective and consider questions such as:

● Does SAP HANA read data from multiple tables when only one is required?
● Does SAP HANA read all records when only one is required?
● Does SAP HANA start with one table even though another table has a much smaller result set?

To gain the insight you need to answer such questions, the SAP HANA studio provides the following features
for query plan analysis:

● Plan explanation
● Plan visualization

4.3.1 Analyzing SQL Execution with the Plan Explanation

You can generate a plan explanation for any SQL statement in the SQL console. You can use this to evaluate
the execution plan that the SAP HANA database follows to execute an SQL statement.

Procedure

1. Enter a query into the SQL console and choose Explain Plan in the context menu.

Note
You can enter multiple statements, separated by the configured separator character (usually a
semicolon), to generate several plan explanations at once.

The plan explanation is displayed on the Result tab.

2. Optional: Run the same statement on different systems/users by changing the SQL connection.
That is, assuming that the tables and views exist in the other systems and you have authorization to
access them.
The plan explanation is displayed on the Result tab.

Results

Plan explanations are also stored in the EXPLAIN_PLAN_TABLE view of the SYS schema for later examination.

SAP HANA Troubleshooting and Performance Analysis Guide

Tools and Tracing PUBLIC 157
4.3.2 Analyzing SQL Execution with the Plan Visualizer
To help you understand and analyze the execution plan of an SQL statement, you can generate a graphical
view of the plan.

Procedure

1. Visualize the plan of the SQL statement in one of the following ways:
a. Enter the statement in the SQL console and choose Visualize Plan in the context menu.
b. On the SQL Plan Cache tab or the Expensive Statements Trace tab of the Performance tab, right-click
the statement and choose Visualize Plan.
A graphical representation of the query, with estimated performance, is displayed.
Figure 37: Visualized Plan

SAP HANA Troubleshooting and Performance Analysis Guide

158 PUBLIC Tools and Tracing
2. Validate the estimated performance by choosing Execute Plan in the context menu. Or use the shortcut
key F8
Another similar high-level graphic is generated with execution time information for each of the parts.

As of SPS 9, by default, a gray color indicates that the operator is not physical, meaning the operator
simply exists to give a more structured graph display.

Almost all of these non-physical operators can be removed from the graph if you prefer, by selecting
'None' for the 'Node Grouping' option provided on the left of the graph.

Note: You can change the colors for each operator type under Window Preferences SAP HANA
PlanViz Graph Appearance .
Figure 38: Executed Plan

SAP HANA Troubleshooting and Performance Analysis Guide

Tools and Tracing PUBLIC 159
 SAP HANA Troubleshooting and Performance Analysis Guide
160 PUBLIC Tools and Tracing
 Note
Execution time is given as a pair of values: "Exclusive" (the execution time of the node), and "Inclusive"
(the execution time including the descendent nodes.

Results

This graphic is a very powerful tool for studying performance of queries on SAP HANA databases. You can
explore the graphic further, for example, you can expand, collapse, or rearrange nodes on the screen. You can
also save the graphic as an image or XML file, for example, so you can submit it as part of a support query.

4.3.2.1 Overview Page

Visualization of execution plans will automatically display an 'Overview' page

Starting from SPS 09 (client version), visualization of execution plans will automatically display an 'Overview'
page as you can see in the screenshot below. Some important KPIs required to begin a performance analysis
are provided so that you can first get a big picture of what is going on before going into the complex details.
Figure 39: Overview Page

The following table describes the nature of each KPI:

Table 42:
KPI Description Comment

Compilation Initial compilation time

SAP HANA Troubleshooting and Performance Analysis Guide

Tools and Tracing PUBLIC 161
KPI Description Comment

Execution Total duration of the query excluding

compilation time

Dominant Operators Operators sorted by their execution You can click on any operator name to
time (top 3) move to the corresponding visualized
operator in the graph

Number of Nodes Number of servers involved in the exe­ You can click on the value to see how
cution much time was spent in each node

Number of Network Transfers Total number of network transfers that You can click on the value to open 'Net­
have occurred during execution work View' which visualizes the trans­
fers in more detail

Dominant Network Transfers Network transfers sorted by their exe­ You can click on any network transfer
cution time (top 3)
to move to the corresponding visual­
ized transfer in the graph.

Note: This section only appears if the

query was executed in a distributed en­
vironment.

SQL Query The statement that was executed

System The system where the execution occur­

red (that is, where final results are
fetched)

Memory Allocated Total memory allocated for executing

the statement

Number of Tables Used Total number of tables touched upon You can click on the value to open the
by any operator during execution 'Tables Used View' for more detail re­
garding each table

Maximum Rows Processed The largest number of rows processed

by any single operator

Result Record Count The final result record count

4.3.2.2 Statement Statistics

Visualization of execution plans for procedures displays a set of statistics for each SQL statement involved in a
procedure.

Context

This set of statistics is automatically visualized when a procedure is executed and it provides a good starting
point for analyzing performance of procedures as it lets you easily drill-down into the most expensive SQL
statements. Basic information such as execution count, execution/compile time, allocated memory size and

SAP HANA Troubleshooting and Performance Analysis Guide

162 PUBLIC Tools and Tracing
so on are provided for each statement so that you can sort the column (criterion) of interest to find the most
expensive statement.

The following pieces of information are available: SQL Statement, Line Number, Execution Count, Execution
Times, Compilation Times, Memory Allocated, Result Record Count, Explain Plan Result, and Procedure
Comment

Figure 40: Statement Statistics view

By right-clicking on any element of the SQL statement, you will have access to these menus:

● Show Plan in Graph: Displays the execution plan corresponding to the selected statement in the context of
entire plan.
● Show Plan in New Window: Displays the execution plan corresponding to the selected statement ONLY in
a separate pop-up window.
● Execute Statement: Enables users to reproduce the execution of a single statement

Note

To use this feature, the 'Keep Temp Tables' button ( ) located on the top-right corner of a graph
editor (Plan Visualizer) must be toggled before obtaining the executed plan of the procedure under
analysis.

Procedure

1. Open an SQL Console

2. Enter any procedure into the console and choose Visualize Plan from the context menu.

SAP HANA Troubleshooting and Performance Analysis Guide

Tools and Tracing PUBLIC 163
3. Choose F8 to see executed plan from the visualized plan.

Statement Statistics will be shown automatically and it will list all the statements that were involved in the
procedure. You can use this view to drill down to the potential root-cause of a long running procedure. For
example you might detect a single statement that takes up most of the time and check whether this is
indeed expected or notice that an identical statement is repeatedly executed with varying statistics (for
example, big gap between minimum execution time and maximum execution time).

4.3.2.3 Timeline View

The Timeline view provides a complete overview of the execution plan based on visualization of sequential
time-stamps.

Context

The following pieces of information are available in the view:

● X-axis: time elapsed since query execution,

● Y-axis: list of operators
● Duration of each operator execution represented by rectangular bars
● The nature of each time-stamp (for example, open, fetch, close, and so on.)

The view supports:

● Synchronized selection with operators in graphical plan

● Hierarchical display of operators based on parent-child relationship
● Re-configuration of scales via a toolbar option

Procedure

1. To see a temporal breakdown of the individual operations processed in the execution of the query, open
the Timeline view.

From the main menu choose Window Show View Timeline .

Figure 41: Timeline View

SAP HANA Troubleshooting and Performance Analysis Guide

164 PUBLIC Tools and Tracing
2. Use this view to drill down to the potential root-cause of a long running query:

For example you might

○ Detect a single operator that takes up most of the time and check whether this is expected
○ Notice that some operators are executed sequentially instead of running in parallel, so you might
check the system environment

Next Steps

However, users should take into account the following point and limitations when analyzing the Timeline view:

● Any gaps or multiple occurrence of 'open', 'fetch', and so on in the Column Search operator invoked by the
row engine indicates that there was another operation called by the column engine in the meantime.
● The adjustment for exclusivity of time-stamps as described above is not applicable for column engine
operators, meaning some operators may not be actually running in parallel but simply calling one another.
● Logical grouping operators for the column engine (those colored in gray) simply display the aggregated
time-stamps of their constituent operators.

4.3.2.4 Operator List for Plan Visualizer

The Operator List view is used within the context of the Plan Visualizer perspective. It lists detailed
characteristics of all operators within a current plan, both visualized and executed.

The Operator List can be used to dynamically explore the operator set along user defined filters in order to
pinpoint specific operators of interest. The view supports:

● Display of various KPIs, for example,. isPhysical (meaning whether an operator is a real, physically
executed one), offset, execution time, CPU time
● Setting of filters along all the columns and KPIs
● Display of the number of operators within the filtered set
● Immediate aggregated information (max, min, sum, and so on) regarding the same KPIs on the filterd
operator set and the remaining set (not within the filter)

SAP HANA Troubleshooting and Performance Analysis Guide

Tools and Tracing PUBLIC 165
● Detailed display of all operators within the filtered set (which can be further sorted)
● Export of the (filtered) operator list to a CSV file
● Forwards and backwards browsing through the history of applied filters.

As of SPS 09 the following features are supported:

● Show or hide columns using the preferences dialog (preferences icon next to save button)
● Change the order of columns using drag and drop
● Order and column visibility are stored in users workspace
● Remote search functionality. Press Ctrl+F when displaying an executed plan and specify search
parameters. When you carry out a search the operator list is updated accordingly
Figure 42: Operator List

You can use the Operator List view to analyze the set of operators within a plan for the occurrence of specific
conditions, even before looking into the visualized plan. For example, you might

1. Filter all operators that process a certain number of (input) rows

2. Further restrict the filter to physical operators (using "x" as the search criteria in filter column "Physical")
3. Display the executed plan, press Ctrl+F and set "Execution Time - At least" to 50
4. Finally, double click on an operator you are interested in to check its positioning within a visualized plan.

4.3.2.5 Network View

The Network View can visualize sequential network data transfers between different servers based on their
given timestamps when an execution plan is based on a distributed query.

Context

The following pieces of information are available in the view:

● X-axis: servers involved,

● Y-axis: time elapsed since query execution
● Duration of each operator represented by rectangular bars
● Size of data transferred in bytes

The view supports:

SAP HANA Troubleshooting and Performance Analysis Guide

166 PUBLIC Tools and Tracing
● Synchronized selection with network operators in graphical plan
● Hover information for network bars
● Zoom in and out

You may use this view to discover any issues related to a distributed environment.

Procedure

1. Open the view

2. Check the results
For example you might notice that an unusually long time is spent on a particular network data transfer, so
you might check the data size and/or network condition and see that network data transfers occur too
frequently between a particular set of servers, so you might improve the distribution set-up.
3. Optimize your distributed landscape based on the results.

SAP HANA Troubleshooting and Performance Analysis Guide

Tools and Tracing PUBLIC 167
4.3.2.6 Tables Used

The table usage view provides an overview on which tables have been used during the processing of a
statement.

Context

The Tables Used view can be used to cross-check your business understanding about which tables are needed
to fulfill a given SQL statement.

The view displays 3 metrics per table:

● Maximum number of entries processed, that is the overall output cardinality of any processing step on
that table in the statement execution.
● Number of accesses, meaning how often a table has been accessed during statement execution
● Maximum processing time, that is the maximum processing time across the possibly multiple table
accesses

The view content can be sorted along any column; double-clicking on any row (table) leads to a corresponding
operator list filtered for physical operators accessing the selected table.

You may use this view to understand whether the data processing of your statement matches your business
expectations.

Procedure

1. Review the tables used in your SQL.

You may detect a table that has been processed which should not be needed from a business perspective,
so you might modify you SQL in a way that this table is no more used.
2. Review the number of times a table is accessed.

SAP HANA Troubleshooting and Performance Analysis Guide

168 PUBLIC Tools and Tracing
 You may see that the number of processed entries is far higher than expected, so you can check if it is
possible to improve the filtering in your SQL statement.

4.3.2.7 Performance Trace

The Performance Trace view displays the list of operators that have performance trace data. The data can be
viewed by selecting any of the operator shown in the list.

Context

The view consists of two main sections:

● The left panel shows list of operators you can select to view performance trace data
● The right panel shows the performance trace data for the operator you have selected

The view supports:

● synchronized selection between operators visualized in the graph

● double-clicking on a particular data to view its full-text

Note
This view is mainly targeted towards advanced users who are well acquainted with the core details of SAP
HANA.

SAP HANA Troubleshooting and Performance Analysis Guide

Tools and Tracing PUBLIC 169
Procedure

1. Look for any additional properties relevant to a particular operator that you cannot find from other views.
2. Fill in missing gaps in the Timeline view by deducing from the sequential list of activities.
3. Further break-down a specific operation into core engine-level details.

4.3.2.8 Plan Trace

Plan trace enables you to collect SQL queries and their execution plans, executed in a given time frame for a
particular application session.

Context

You can access the plan trace configuration wizard in SAP HANA studio in two ways:

● By right-clicking on a particular session you wish to trace in the Administration editor -> Performance ->
Sessions
● Going to Administration editor -> Trace Configuration tab

For each SQL query that has been traced, you will be able to visualize the execution plan for performance
analysis.

Once you stop an active trace, a Plan Trace Editor will be automatically opened displaying the queries that
have been traced.

As of SPS 09, only 'SELECT' statements are traced with Plan trace. You can double-click a particular
statement or open "Visualize Plan" from the context menu to get the execution plan.

SAP HANA Troubleshooting and Performance Analysis Guide

170 PUBLIC Tools and Tracing
Procedure

1. Open the SQL Console

2. Open Administration editor Performance tab Sessions tab

3. Right-click on the session which corresponds to the SQL Console that you opened in step 1.
The column Application Source will contain the keyword 'SQLExecuteFormEditor'
4. From the context menu, choose Configure Trace Plan Trace
5. 'Press 'Finish' in the pop-up configuration wizard
6. Execute some SELECT queries in the SQL Console
Only SELECT queries are supported with SPS 09.
7. After execution, go back to the sessions list and deactivate 'Plan Trace' via configuration wizard (as in
steps 4 to 6)
8. An editor (Plan Trace editor) will open automatically showing the list of queries that you have executed in
'SQL Console'
9. Each item (query) can be double-clicked for visualization of its execution plan

SAP HANA Troubleshooting and Performance Analysis Guide

Tools and Tracing PUBLIC 171
4.3.2.9 Calculation Scenario Visualization

By navigating to a particular calculation view in the Systems view, you can access the Visualize View context
menu, which will visualize the calculation scenario associated with the view.

The calculation scenario visualizer consists of following three main sections:

● left: an overview which provides a list of calculation nodes involved in the scenario with their main
properties
● middle: the visualized graph of calculation scenario
● right: an input tab which provides mapping information of the selected node, an output tab which provides
attributes of a selected node, and a source tab which shows the raw JSON data of the selected node

The visualizer supports:

● link with 'Properties' view for each selected node

● double-clicking on a particular data source to drill down to its calculation scenario

You may use this visualizer to check whether the scenario is designed according to your understanding, and, if
not, find potential tweaking points.

SAP HANA Troubleshooting and Performance Analysis Guide

172 PUBLIC Tools and Tracing
4.4 Static Cached Views

Complex SQL views and calculation views place a heavy load on the system. You may be able to use cached
views to reduce CPU consumption (and thus to increase overall system throughput) or to reduce the response
time of queries.

Frequently accessed views may consume an excessive amount of CPU. Cached views can trade this CPU
consumption for an increased memory consumption by caching the result of evaluating a view and reusing this
cached data when the view is accessed by a query.

Using cached views the query result of a view is cached in memory and refreshed only periodically. On the one
hand, this avoids having to reevaluate the view each time the view is accessed, on the other hand, the cached
data might be stale if the tables used by the view are updated after the last refresh. However, the data
retention period which determines how old the data may become is configurable and the age of the cached
data is also returned to the database client.

Cached views must be implemented in combination with careful testing to validate that correct results are
returned, that the cached views is actually used, and that indeed the CPU consumption is reduced. Tools
which can help with this are available in SAP Studio such as Explain Plan and Visualize Plan.

Scope

Generally caching can be used for calculation views, SQL views, table functions which return the result of a
single SQL SELECT statement, and CDS views (which are translated into table functions). To save memory,
cached views store aggregated values (aggregation types SUM, MIN, MAX, and COUNT are supported).

Not all views may be suitable for caching (for example views using time functions which depend upon the
current time would not be suitable). Caching cannot be used for scripted views, join views, OLAP views, and
hierarchy views. Other limitations may apply, for example, if analytic privileges have been applied to the data.

Configuring the Memory Budget

A maximum cache size (a memory budget) must be defined to manage the maximum amount of memory
available to cache cached views. This is configured in the indexserver.ini file in the total_size parameter in
the result_cache section of the file.

Parameter Default Detail

total_size 10000 MB The memory budget in MB available for cached views.

If the memory consumed for caching views exceeds this threshold an alert is raised and, where necessary,
cached views will be dropped from memory resulting in higher response times for previously cached data. In
this case administrators may increase the memory budget size or create capacity in the budget by, for
example, dropping cached views which are rarely used.

SAP HANA Troubleshooting and Performance Analysis Guide

Tools and Tracing PUBLIC 173
If a view fails to be created then it is added to an exclusion list of views which have not been successfully
cached (view M_VIEW_CACHE_EXCLUSIONS). There are two common cases where this occurs:

● If the time to calculate the result of the view is longer than the retention time.
● If the memory required to cache the result of the view is larger than the memory budget.

The following example query selects basic details of failed cached views:

SELECT SCHEMA_NAME, VIEW_NAME, REASON, EXCLUDE_TIME

FROM M_RESULT_CACHE_EXCLUSIONS

4.4.1 Creating and Using Cached Views

This topic illustrates how cached views are created based on an example SQL view.

To save memory cached SQL views are based on aggregated results (SUM, MIN, MAX and COUNT are
supported); only the aggregated values are stored. The SQL optimizer uses the aggregation type of each
column in the top-most select statement as the basis for the cached result.

When a user submits a query to a cached view a matching process analyzes the request and the definition of
the query to see if the cached data can be used or if the request must read the data from the database.

To exploit the cached data the query must match with the definition of the result cache and a number of
matching rules apply:

● The query must reference the same columns (or a subset of the columns) in the cache.
● The query must explicitly reference the view or a sub view referenced by the view.
● The predicate in the query must be the same as that used in creating the view, any additional conditions
which are added must be more restrictive filters.

Examples

Aggregation
This example creates a view and then uses the alter view syntax to add the view to the cache with a retention
period of 120 minutes.

CREATE VIEW SIMPLE_VIEW AS

(SELECT A, SUM(KF1) AS KF1, MIN(KF2) AS KF2, MAX(KF3) AS KF3
FROM SIMPLE_TABLE GROUP BY A)

ALTER VIEW SIMPLE_VIEW ADD CACHE RETENTION 120 OF A, KF1, KF2, KF3, KF4;

In the following example queries the first two statements consistently use the same aggregated values as
defined in the query and can exploit the cache. The third example cannot use the cached data because it
requests unaggregated details which are not included in the cache:

SELECT SUM(KF1) FROM SIMPLE_VIEW;

SELECT SUM(KF1), MIN(KF2), MAX(KF3) FROM SIMPLE_VIEW GROUP BY A;

SAP HANA Troubleshooting and Performance Analysis Guide

174 PUBLIC Tools and Tracing
 SELECT KF1, KF2, KF3 FROM SIMPLE_VIEW;

Filters
This second example illustrates filtering. In this case the column B is also cached with an additional filter:

ALTER VIEW SIMPLE_VIEW ADD CACHE RETENTION 120 OF A, KF1, KF2, KF3, KF4 FILTER B
> 3;

The following example queries both apply additional more restrictive filters and because the selections are
within the scope of the cached view both queries can use the cache:

SELECT SUM(KF1) FROM SIMPLE_VIEW WHERE B > 3 AND B < 10;

SELECT SUM(KF1) FROM SIMPLE_VIEW WHERE B > 3 AND A = 1;

Using OF to Define a Projection List

The cached view examples given above use the OF keyword to define a projection list. This reduces the
amount of cached data as only rows that match the filter condition (the named columns) are cached. Queries
which request data from other columns cannot exploit the cache.

When you use filters in conjunction with the OF keyword it may be necessary to select and cache more
columns than explicitly specified in the ALTER VIEW statement. This is shown in the following example where a
filter is applied on a column which is not part of the projection list:

ALTER VIEW "CALC_VIEW" ADD CACHE RETENTION 100 OF "A" FILTER "B" = 100

In this case it is necessary to cache both A and B to ensure that the correct results can be returned.

Cached Views and Analytic Privileges

Analytic privileges may be applied to views to restrict access to data which a user is not authorized to see (this
may be done, for example, on the basis of a user's geographical area or organization unit). Normally, only
views with static (design time) XML privileges can be cached. If you need to use other types of analytic
privileges (SQL or run-time privileges) then it is possible to configure cached views to cache all data before the
restriction is applied and apply the restriction when the data is queried. To do this you must use an additional
configuration parameter and a hint as described here.

This feature is enabled using the following configuration setting in the result_cache section of the
indexserver.ini file:

Parameter Default Detail

before_analytic_privilege False Set this to True to enable the option of caching the result before
applying analytic privileges.

SAP HANA Troubleshooting and Performance Analysis Guide

Tools and Tracing PUBLIC 175
Once the feature is enabled the following hints are supported:

● HINT(RESULT_CACHE_BEFORE_ANALYTIC_PRIVILEGE) - this hint builds the cache before the

authorization restrictions are applied (that is, read all the data and apply the restrictions when the data is
queried so that only authorized data is selected for each user).
● HINT(RESULT_CACHE_AFTER_ANALYTIC_PRIVILEGE)- this hint builds the cache in the normal way by
applying analytic privileges first. This option is more efficient in terms of memory usage since less data is
cached but is only suitable for views where static XML privileges have been applied.

Other Hints

A number of hints are available for use with cached views:

● HINT (IGNORE_RESULT_CACHE) Use this hint in a query to by-pass the cache.

● HINT(NO_RESULT_CACHE)
● HINT(RESULT_CACHE)
● HINT(RESULT_CACHE_MAX_LAG(seconds))
● HINT(RESULT_CACHE_NON_TRANSACTIONAL) Allows join/union to avoid transactional inconsistency.
● HINT(RESULT_CACHE_NO_REFRESH) Access existing cached data without refreshing it even if its
retention period is over.

Result Cache HDI Plugin

To transform and deploy design time DDL-based result cache definitions the result cache plugin is available.
The following example defines a result cache for the view named 'A_Sql_View':

RESULT CACHE "_SYS_CACHE#sap::A_Sql_View" ON VIEW "sap::A_Sql_View" WITH

RETENTION 30

To cache a table function use ON FUNCTION instead of ON VIEW. Any values after the keyword WITH are
copied to the corresponding SQL statement, so other parameters such as defining a projection list, a filter or
the force flag are also supported.

Refer to the SAP HANA Developer Guide For SAP HANA XS Advanced Model for further details.

Related Information

SAP HANA Developer Guide (For SAP HANA XS Advanced Model)

SAP HANA Troubleshooting and Performance Analysis Guide

176 PUBLIC Tools and Tracing
4.4.2 Resources for Testing Cached Views
This topic uses an example calculation view to illustrate how you can work with tools to test cached views.

After caching a view it is important to verify that the view has been successfully cached, that correct results
are returned, that the cached views is actually used and is effectively reducing CPU consumption. The
following example uses the ALTER VIEW syntax to add a calculation view result to the cache with a retention
period of 60 minutes:

ALTER VIEW "_SYS_BIC"."sap.hba.ecc/CostCenterPlanActualCommitmentQuery" ADD

CACHE RETENTION 60;

Note that the top-most node of a calculation view should be an aggregation type node. If it is not but the
calculation view does implicit grouping then caching will fail.

Carry out the following checks:

1) Check the HAS_CACHE value for the view:

select schema_name, view_name, has_cache

from views
where view_name = 'my_view';

Any value for HAS_CACHE other than 'none' indicates the successful creation of the cached view.

2) After running a query against the cached view, use the Explain Plan tool (context menu in the SQL Console)
to confirm the source of the query result. The OPERATOR_NAME value should show RESULT_CACHE.

3) Continue to test the cached view by monitoring the execution times of queries based on the view. Expect
the first execution of the query to be slow and look for performance improvements on subsequent runs. If
there is no reduction in response time or CPU consumption when the cached view is used then it is not
beneficial.

4) Use monitoring views M_RESULT_CACHE and M_HEAP_MEMORY to see more details of the cached view
such as memory consumption and data refresh details.

4.5 Advanced Analysis

If have you an advanced knowledge of SAP HANA and SQL databases and you suspect that automated
processes are making poor decisions that have negative impacts on query performance, you can perform
advanced analyses to better understand how those decisions were made.

In SAP HANA, you can use specific SQL commands to produce trace information to help analyze the following
processes:

● Table joins
● Column searches

Recommendation
Perform these types of analysis only if analyses of query plans and SQL statements were not enough to find
the root cause of slow query performance.

SAP HANA Troubleshooting and Performance Analysis Guide

Tools and Tracing PUBLIC 177
4.5.1 Analyzing Column Searches

In SAP HANA, if a column search takes a long time, you can analyze how the query-optimizer performed the
column search. A query-optimizer (qo) trace of a single SAP HANA table search provides the details you need
for such an analysis.

Context

The qo trace provides a lot of information that is hard to consume if you are not an SAP HANA query-optimizer
expert; however, it does provide some useful information for performance analysis. From the information
within the trace files, you can see which column the query-optimizer decided to use as the first column in the
column search and you can determine whether that decision negatively impacted the performance of the
column search.

To start a user-specific qo trace and analyze the relevant trace information, proceed as follows.

Procedure

1. In the Administration editor, choose Trace Configuration and create a new user-specific trace.
The Trace Configuration dialog box opens.
2. Specify a context name.
The context name appears as part of the trace file name and should be easy for you to recognize and later
find.
3. Specify your database user or application user.
4. Select the Show All Components checkbox.
5. Enter qo as filter text and search for the trex_qo component.
6. For the trex_qo component, select DEBUG as the system trace level.
7. Choose Finish.
8. Run the query you want to trace.
9. Switch off the trace by deleting the user-specific trace configuration.
10. On the Diagnosis Files tab, search for the trace file, open the file, and select Show Entire File.
11. In the trace section, analyze the trace information for each term (WHERE condition).
a. Find the sections detailing the estimated results for the terms.
These sections are marked with GetEstimation.cpp.
b. Find the sections detailing the actual search results for the terms.
These sections are marked with Evaluate.cpp.
c. Compare the estimated results with the actual search results.
The query-optimizer selects which column to use as the first column of the search based on the term
with the lowest estimated number of results.

SAP HANA Troubleshooting and Performance Analysis Guide

178 PUBLIC Tools and Tracing
Results

If the actual results indicate that a different term should have been used to start the column search, then this
may represent the source of poor performance. For more detailed analysis, you can send the trace file to SAP
Support.

4.5.2 Analyzing Table Joins

In SAP HANA, if a query on joined tables takes a long time, you can analyze how the tables are joined and in
what order. A join evaluation (je) trace of joined SAP HANA tables provides the details you need for such an
analysis.

Context

The je trace provides a lot of information that is hard to consume if you are not an SAP HANA join engine
expert; however, it does provide some useful information for performance analysis. From the information
within the trace files, you can see which table is used as the first table when processing a join and how the
order of tables in the join is defined. You can use this information to determine whether query performance is
negatively impacted by the table join.

To start a je trace and analyze the relevant trace information, proceed as follows:

Procedure

1. In the Administration editor, choose Trace Configuration and create a new user-specific trace.
The Trace Configuration dialog box opens.
2. Specify a context name.
The context name appears as part of the trace file name and should be easy for you to recognize and later
find.
3. Specify your database user or application user.
4. Select the Show All Components checkbox.
5. Enter join as filter text and search for the join_eval component.
6. For the join_eval component, select DEBUG as the system trace level.
7. Choose Finish.
8. Run the query you want to trace.
9. Switch off the trace by deleting the user-specific trace configuration.
10. On the Diagnosis Files tab, search for the indexserver trace file, open the file, and select Show Entire File.
11. From the end of the file, search backwards for the beginning of the trace section.
The trace section starts with i TraceContext TraceContext.cpp.

SAP HANA Troubleshooting and Performance Analysis Guide

Tools and Tracing PUBLIC 179
12. In the trace section, analyze the following trace information:
○ Estimations for the WHERE conditions
○ Table size and join conditions
○ Join decision

4.6 Additional Analysis Tools for Support

To complement the standard tools for performance analysis, SAP HANA provides additional analysis tools that
SAP Support can use to help determine the cause of performance issues.

The following analysis tools are available in SAP HANA; however, these tools are intended only for use when
requested by SAP Support:

● Performance trace
This tool records performance indicators for individual query processing steps in database kernel.
● Kernel profiler
This tool provides information about hotspots and expensive execution paths during query processing.

4.6.1 Performance Trace

The performance trace is a performance tracing tool built into the SAP HANA database. It records
performance indicators for individual query processing steps in the database kernel. It is inactive by default.

Information collected includes the processing time required in a particular step, the data size read and written,
network communication, and information specific to the operator or processing-step-specific (for example,
number of records used as input and output). The performance trace can be enabled for multiple tenant
databases at the same time to analyze cross-database queries.

It is recommended that you start performance tracing immediately before running the command(s) that you
want to analyze and stop it immediately after they have finished. When you stop tracing, the results are saved
to trace files that you can access on the Diagnosis Files tab of the Administration editor. You cannot analyze
these files meaningfully in the SAP HANA studio, but instead must use a tool capable of reading the output
format (*.tpt). SAP Support has tools for evaluating performance traces.

You activate and configure the performance trace in the Administration editor on the Trace Configuration tab.

Related Information

Performance Trace Options [page 181]

SAP HANA Troubleshooting and Performance Analysis Guide

180 PUBLIC Tools and Tracing
4.6.2 Performance Trace Options

Several options are available for configuring the performance trace.

Standard Trace Mode

Table 43:
Option Description

Trace file The file to which the trace data is automatically saved after the performance trace is stop­
ped

User and application filters Filters to restrict the trace to a single specific database user, a single specific application
user, and a single specific application

Passport trace level If you are activating the performance trace as part of an end-to-end trace scenario with the
Process Monitoring Infrastructure (PMI), you can specify the passport trace level as an ad­
ditional filter. This means that only requests that are marked with a passport of the speci­
fied level are traced.

For more information see, SAP Library for NetWeaver on SAP Help Portal.

Note
Process tracing is possible only for components in the ABAP and Business Objects
stacks.

Trace execution plans You can trace execution plans in addition to the default trace data.

Function profiler The function profiler is a very fine-grained performance tracing tool based on source code
instrumentation. It complements the performance trace by providing even more detailed
information about the individual processing steps that are done in the database kernel.

Duration How long you want tracing to run

If a certain scenario is to be traced, ensure that you enter a value greater than the time it
takes the scenario to run. If there is no specific scenario to trace but instead general system
performance, then enter a reasonable value. After the specified duration, the trace stops
automatically.

Related Information

Process Monitoring with PMI (Process Monitoring Infrastructure)

SAP HANA Troubleshooting and Performance Analysis Guide

Tools and Tracing PUBLIC 181
4.6.3 Kernel Profiler

The kernel profiler is a sampling profiler built into the SAP HANA database. It can be used to analyze
performance issues with systems on which third-party software cannot be installed, or parts of the database
that are not accessible by the performance trace. It is inactive by default.

Caution
To be able to use the kernel profile, you must have the SAP_INTERNAL_HANA_SUPPORT role. This role is
intended only for SAP HANA development support.

The kernel profile collects, for example, information about frequent and/or expensive execution paths during
query processing.

It is recommended that you start kernel profiler tracing immediately before you execute the statements you
want to analyze and stop it immediately after they have finished. This avoids the unnecessary recording of
irrelevant statements. It is also advisable as this kind of tracing can negatively impact performance. When you
stop tracing, the results are saved to trace files that you can access on the Diagnosis Files tab of the
Administration editor. You cannot analyze these files meaningfully in the SAP HANA studio, but instead must
use a tool capable of reading the configured output format, that is KCacheGrind or DOT (default format).

You activate and configure the kernel profile in the Administration editor on the Trace Configuration tab.

Related Information

Kernel Profiler Options [page 182]

4.6.4 Kernel Profiler Options

Several options are available for configuring the kernel profiler.

Table 44:
Option Description

Service(s) to profile The service(s) that you want to profile

Wait time The amount of time the kernel profiler is to wait between call stack retrievals

When you activate the kernel profiler, it retrieves the call stacks of relevant threads several
times. If a wait time is specified, it must wait the specified time minus the time the previous
retrieval took.

SAP HANA Troubleshooting and Performance Analysis Guide

182 PUBLIC Tools and Tracing
 Option Description

Memory limit Memory limit that will stop tracing

The kernel profiler can potentially use a lot a memory. To prevent the SAP HANA database
from running out of memory due to profiling, you can specify a memory limit that cannot be
exceeded.

Database user, application The database user and/or application user you want to profile
user

Use KCachegrind format to Output format of trace files (configurable when you stop tracing)
write trace files

4.6.5 Diagnosis Information

You can collect diagnosis information in the SAP HANA studio and using command line scripts.

To collect this information, use the SAP HANA studio Administration Editor, navigate to Diagnosis Files
Diagnosis Information and use the Collect function.

The SQL variant can be used when SAP HANA is online, otherwise choose the Python script.

4.7 Analysis Tools in SAP HANA Web-based Developer

Workbench

There are a number of tools available for application performance analysis in the SAP HANA Web-based
Developer Workbench.

SQL Console

The extended SQL Console (part of the Catalog perspective) supports implicit and explicit performance
measurement while executing a SQL statement. It allows you to easily acquire sound measurement data and
to assess whether a SQL statement is problematic.

SAP HANA Troubleshooting and Performance Analysis Guide

Tools and Tracing PUBLIC 183
Implicit performance measurement can be triggered via F9 run command and provides execution time
information for the database, XS, and UI layers. Additional information to judge the measurement quality like
table locks and system alerts is also provided. Detailed performance analysis supports the repeated execution
of statements thus allows you to check whether performance characteristics are stable.

You can use this feature to quickly check SQL performance and to determine whether the observed poor
performance is caused by system load or variability.

Expensive Statements Trace

The expensive statements trace in the Trace perspective allows you to quickly analyze how the SQL layer of
your SAP HANA system is used (globally or within a specific session/application). It allows you to analyze if
the SAP HANA database is used effectively and efficiently for a given application.

SAP HANA Troubleshooting and Performance Analysis Guide

184 PUBLIC Tools and Tracing
The expensive statement view supports filtering for passport-based application sessions - if the trace is
configured accordingly by the system administrator and you run your application with SAP passport. For each
statement you can see metrics such as start time, number of processed records and many more.

Furthermore, a statistical summary is provided that gives insight on the overall SQL load. It details metrics
such as:

● Number of statements executed

● Number of unique statements executed
● Number of result records

Last, for each unique statement summary statistics are shown that detail the frequency of their usage as well
as aggregated duration metrics.

You can use the expensive statement trace to analyze:

● the overall SQL load in your current system

● the SQL interaction of a specific application, for example, to figure out if an application that is not currently
performing well due to an issue on the application layer or on the database layer.

Immediate Database Performance Feedback

This feature is part of the Immediate Feedback context in WebIDE.It provides a performance overview of all
SQL statements which are executed during an Immediate Feedback session, thus helps application developers
to immediately understand the number and performance of their DB-related statements.

SAP HANA Troubleshooting and Performance Analysis Guide

Tools and Tracing PUBLIC 185
The Database Performance overview provides 3 metrics for each SQL statement: step, statement and
execution.

● Step: the actual immediate feedback step in which a SQL statement was executed
● Statement: the SQL statement
● Execution time: the execution time of the statement (in milliseconds)

You can use this feature to understand the performance behavior of the SQL statements (for example, finding
expensive statements that belong to a particular function) and exploit this understanding for possible further
performance investigations, like subsequently taking a dedicated measurement using the SQL Console.

SAP HANA Troubleshooting and Performance Analysis Guide

186 PUBLIC Tools and Tracing
5 Reference: Alerts

Use this reference section to look up Alerts, get specific information on how to handle the cause, and identify
any additional information.

Table 45: Alerts

ID Name Description User Action Category Further Infor­
mation

0 Internal statis­ Identifies internal statistics server Resolve the problem. For Availabil­ SAP Note
tics server prob­ problem. more information, see the ity 1803039, SAP
lem trace files. You may need to Note: 1803039
activate tracing first.

1 Host physical Determines what percentage of to­ Investigate memory usage of Memory Memory Prob­
memory usage tal physical memory available on processes. lems, SAP Note:
the host is used. All processes con­ 1840954
suming memory are considered,
including non-SAP HANA proc­
esses.

2 Disk usage Determines what percentage of Investigate disk usage of Disk Disk Related,
each disk containing data, log, and processes. Increase disk SAP Note:
trace files is used. This includes space, for example by 1900643
space used by non-SAP HANA shrinking volumes, deleting
files. diagnosis files, or adding ad­
ditional storage.

3 Inactive services Identifies inactive services. Investigate why the service is Availabil­ SAP Note:
inactive, for example, by ity 1902033
checking the service's trace
files.

4 Restarted serv­ Identifies services that have re­ Investigate why the service Availabil­ SAP Note:
ices started since the last time the had to restart or be re­ ity 1909660
check was performed. started, for example, by
checking the service's trace
files.

5 Host CPU Usage Determines the percentage CPU Investigate CPU usage. CPU CPU Related,
idle time on the host and therefore SAP Note:
whether or not CPU resources are 1909670
running low.

10 Delta merge Determines whether or not the 'ac­ Change in SYSTEM layer the Configu­ Delta Merge ,
(mergedog) tive' parameter in the 'mergedog' parameter active in sec­ ration SAP Note:
configuration section of system configuration tion(s) mergedog to yes 1909641
file(s) is 'yes'. mergedog is the sys­
tem process that periodically
checks column tables to determine
whether or not a delta merge oper­
ation needs to be executed.

SAP HANA Troubleshooting and Performance Analysis Guide

Reference: Alerts PUBLIC 187
ID Name Description User Action Category Further Infor­
mation

12 Memory usage Determines what percentage of al­ Increase the shared memory Memory Memory Prob­
of name server located shared memory is being size of the name server. In lems, SAP Note:
used by the name server on a host. the 'topology' section of the 1977101
nameserver.ini file, increase
the value of the 'size' param­
eter.

16 Lock wait time­ Determines whether the In the 'transaction' section of Configu­ Transactional
out configura­ 'lock_wait_timeout' parameter in the indexserver.ini file, set ration Problems, SAP
tion the 'transaction' section of the in­ the 'lock_wait_timeout' pa­ Note: 1909707
dexserver.ini file is between rameter to a value between
100,000 and 7,200,000. 100,000 and 7,200,000 for
the System layer.

17 Record count of Determines the number of records Consider partitioning the ta­ Memory Memory Prob­
non-partitioned in non-partitioned column-store ta­ ble only if you expect it to lems, SAP
column-store ta­ bles. Current table size is not criti­ grow rapidly. HANA Adminis­
bles cal. Partitioning need only be con­ tration Guide >
sidered if tables are expected to Table Partition­
grow rapidly (a non-partitioned ta­ ing, SAP Note:
ble cannot contain more than 1909763
2,147,483,648 (2 billion) rows).

20 Table growth of Determines the growth rate of non- Consider partitioning the ta­ Memory Memory Prob­
non-partitioned partitioned columns tables. ble. lems, SAP
column-store ta­ HANA Adminis­
bles tration Guide >
Table Partition­
ing, SAP Note:
1910140

21 Internal event Identifies internal database events. Resolve the event and then Availabil­ SAP Note:
mark it as resolved by exe­ ity 1977252
cuting the SQL statement
ALTER SYSTEM SET EVENT
HANDLED '<host>:<port>'
<id>. Note that this is not
necessary for INFO events.

22 Notification of Determines whether or not there Investigate the alerts. Availabil­

all alerts have been any alerts since the last ity
check and if so, sends a summary
e-mail to specified recipients.

23 Notification of Determines whether or not there Investigate the alerts. Availabil­

medium and have been any medium and high ity
high priority priority alerts since the last check
alerts and if so, sends a summary e-mail
to specified recipients.

24 Notification of Determines whether or not there Investigate the alerts. Availabil­

high priority have been any high priority alerts ity
alerts since the last check and if so,
sends a summary e-mail to speci­
fied recipients.

SAP HANA Troubleshooting and Performance Analysis Guide

188 PUBLIC Reference: Alerts
 ID Name Description User Action Category Further Infor­
mation

25 Open connec­ Determines what percentage of the Investigate why the maxi­ Sessions/ SAP Note:
tions maximum number of permitted mum number of permitted Transac­ 1910159
SQL connections are open. The open connections is being tions
maximum number of permitted approached.
connections is configured in the
"session" section of the index­
server.ini file.

26 Unassigned vol­ Identifies volumes that are not as­ Investigate why the volume Configu­ SAP Note:
umes signed a service. is not assigned a service. For ration 1910169
example, the assigned serv­
ice is not active, the removal
of a host failed, or the re­
moval of a service was per­
formed incorrectly.

27 Record count of Determines the number of records Consider repartitioning the Memory Memory Prob­
column-store ta­ in the partitions of column-store table. lems, SAP
ble partitions tables. A table partition cannot HANA Adminis­
contain more than 2,000,000,000 tration Guide >
(2 billion) rows. Table Partition­
ing, SAP Note:
1910188

28 Most recent sa­ Determines how long ago the last Investigate why there was a Disk SAP Note:
vepoint opera­ savepoint was defined, that is, how delay defining the last save­ 1977291
tion long ago a complete, consistent point and consider triggering
image of the database was per­ the operation manually by
sisted to disk. executing the SQL statement
ALTER SYSTEM SAVE­
POINT.

29 Size of delta Determines the size of the delta Investigate the delta merge Memory Delta Merge ,
storage of col­ storage of column tables. history in the monitoring SAP Note:
umn-store ta­ view 1977314
bles M_DELTA_MERGE_STATIS­
TICS. Consider merging the
table delta manually.

30 Check internal Determines whether or not the Resolve the disk-full event as Disk SAP Note:
disk full event disks to which data and log files are follows: In the Administration 1898460
written are full. A disk-full event Editor on the Overview tab,
causes your database to stop and choose the "Disk Full Events"
must be resolved. link and mark the event as
handled. Alternatively, exe­
cute the SQL statements AL­
TER SYSTEM SET EVENT
ACKNOWLEDGED
'<host>:<port>' <id> and AL­
TER SYSTEM SET EVENT
HANDLED '<host>:<port>'
<id>.

SAP HANA Troubleshooting and Performance Analysis Guide

Reference: Alerts PUBLIC 189
ID Name Description User Action Category Further Infor­
mation

31 License expiry Determines how many days until Obtain a valid license and in­ Availabil­ Security, Au­
your license expires. Once your li­ stall it. For the exact expira­ ity thorization and
cense expires, you can no longer tion date, see the monitoring Licensing, SAP
use the system, except to install a view M_LICENSE. Note: 1899480
new license.

32 Log mode LEG­ Determines whether or not the da­ If you need point-in-time re­ Backup Configuration
ACY tabase is running in log mode "leg­ covery, reconfigure the log Parameter Is­
acy". Log mode "legacy" does not mode of your system to sues,
support point-in-recovery and is "normal". In the "persis­
not recommended for productive tence" section of the
systems. global.ini configuration file,
set the parameter
"log_mode" to "normal" for
the System layer. When you
change the log mode, you
must restart the database
system to activate the
changes. It is also recom­
mended that you perform a
full data backup.

33 Log mode Determines whether or not the da­ If you need point-in-time re­ Backup SAP HANA Ad­
OVERWRITE tabase is running in log mode covery, reconfigure the log ministration
"overwrite". Log mode "overwrite" mode of your system to Guide > Backing
does not support point-in-recovery "normal". In the "persis­ up and Recover­
(only recovery to a data backup) tence" section of the ing the SAP
and is not recommended for pro­ global.ini configuration file, HANA Database,
ductive systems. set the parameter
"log_mode" to "normal" for
the System layer. When you
change the log mode, you
must restart the database
system to activate the
changes. It is also recom­
mended that you perform a
full data backup.

34 Unavailable vol­ Determines whether or not all vol­ Investigate why the volume Configu­ SAP HANA Ad­
umes umes are available. is not available. ration ministration
Guide > Backing
up and Recover­
ing the SAP
HANA Database,
SAP Note:
1900682

35 Existence of Determines whether or not a data Perform a data backup as Backup SAP HANA Ad­
data backup backup exists. Without a data soon as possible. ministration
backup, your database cannot be Guide > Backing
recovered. up and Recover­
ing the SAP
HANA Database,
SAP Note:
1900728

SAP HANA Troubleshooting and Performance Analysis Guide

190 PUBLIC Reference: Alerts
 ID Name Description User Action Category Further Infor­
mation

36 Status of most Determines whether or not the Investigate why the last data Backup SAP HANA Ad­
recent data most recent data backup was suc­ backup failed, resolve the ministration
backup cessful. problem, and perform a new Guide > Backing
data backup as soon as pos­ up and Recover­
sible. ing the SAP
HANA Database,
SAP Note:
1900795

37 Age of most re­ Determines the age of the most re­ Perform a data backup as Backup SAP HANA Ad­
cent data cent successful data backup. soon as possible. ministration
backup Guide > Backing
up and Recover­
ing the SAP
HANA Database,
SAP Note:
1900730

38 Status of most Determines whether or not the Investigate why the log Backup SAP HANA Ad­
recent log back­ most recent log backups for serv­ backup failed and resolve the ministration
ups ices and volumes were successful. problem. Guide > Backing
up and Recover­
ing the SAP
HANA Database,
SAP Note:
1900788

39 Long-running Identifies long-running SQL state­ Investigate the statement. Sessions/ Transactional
statements ments. For more information, see Transac­ Problems, SAP
the table _SYS_STATIS­ tions Note: 1977262
††Ò⁄€£)Äšíá'·9Uk#k_˛˝ł‚@ËÌz…ß"¾¿dŠ:ç
NING_STATEMENTS.

40 Total memory Determines what percentage of the Consider partitioning or re­ Memory Memory Prob­
usage of col­ effective allocation limit is being partitioning the table. lems, SAP Note:
umn-store ta­ consumed by individual column- 1977268
bles store tables as a whole (that is, the
cumulative size of all of a table's
columns and internal structures)

41 In-memory Da­ Determines whether or not there is For more information, see Availabil­ SAP Note:
taStore activa­ a problem with the activation of an the table _SYS_STATIS­ ity 1665553, SAP
tion in-memory DataStore object. ††Ò⁄€£)Äšíá(·:Uw#}_˝ł‚PËÆz†ß>¾²tŠ,ç Note: 1977230
TRACTOR_STATUS and SAP
Note 1665553.

42 Long-idling cur­ Identifies long-idling cursors. Close the cursor in the appli­ Sessions/ Transactional
sors cation, or kill the connection Transac­ Problems, SAP
by executing the SQL state­ tions Note: 1900261
ment ALTER SYSTEM DIS­
CONNECT SESSION <LOGI­
CAL_CONNECTION_ID>. For
more information, see the ta­
ble HOST_LONG_IDLE_CUR­
SOR (_SYS_STATISTICS).

SAP HANA Troubleshooting and Performance Analysis Guide

Reference: Alerts PUBLIC 191
ID Name Description User Action Category Further Infor­
mation

43 Memory usage Determines what percentage of its Check for services that con­ Memory Memory Prob­
of services effective allocation limit a service is sume a lot of memory. lems, SAP Note:
using. 1900257

44 Licensed mem­ Determines what percentage of li­ Increase licensed amount of Memory Memory Prob­
ory usage censed memory is used. main memory. You can see lems, SAP Note:
the peak memory allocation 1899511
since installation in the sys­
tem view M_LICENSE (col­
umn PRODUCT_USAGE).

45 Memory usage Determines what percentage of the Consider partitioning or re­ Memory Memory Prob­
of main storage effective allocation limit is being partitioning the table. lems, SAP Note:
of column-store consumed by the main storage of 1977269
tables individual column-store tables.

46 RTEdump files Identifies new runtime dump files Check the contents of the Diagnosis SAP Note:
(*rtedump*) have been generated dump files. Files 1977099
in the trace directory of the sys­
tem. These contain information
about, for example, build, loaded
modules, running threads, CPU,
and so on.

47 Long-running Identifies long-running serializable Close the serializable trans­ Sessions/ Transactional
serializable transactions. action in the application or Transac­ Problems,
transactions kill the connection by execut­ tions
ing the SQL statement AL­
TER SYSTEM DISCONNECT
SESSION <LOGICAL_CON­
NECTION_ID>. For more in­
formation, see the table
HOST_LONG_SERIALIZA­
BLE_TRANSACTION
(_SYS_STATISTICS).

48 Long-running Identifies long-running uncommit­ Close the uncommitted Sessions/ Transactional

uncommitted ted write transactions. transaction in the application Transac­ Problems, SAP
write transac­ or kill the connection by exe­ tions Note: 1977276
tions cuting the SQL statement
ALTER SYSTEM DISCON­
NECT SESSION <LOGI­
CAL_CONNECTION_ID>. For
more information, see the ta­
ble HOST_UNCOMMIT­
TED_WRITE_TRANSACTION
(_SYS_STATISTICS).

49 Long-running Identifies long-running blocking sit­ Investigate the blocking and Sessions/ Transactional
blocking situa­ uations. blocked transactions and if Transac­ Problems, SAP
tions appropriate cancel one of tions Note: 2079396
them.

SAP HANA Troubleshooting and Performance Analysis Guide

192 PUBLIC Reference: Alerts
 ID Name Description User Action Category Further Infor­
mation

50 Number of diag­ Determines the number of diagno­ Investigate the diagnosis Diagnosis SAP Note:
nosis files sis files written by the system (ex­ files. Files 1977162
cluding zip-files). An unusually
large number of files can indicate a
problem with the database (for ex­
ample, problem with trace file rota­
tion or a high number of crashes).

51 Size of diagnosis Identifies large diagnosis files. Un­ Check the diagnosis files in Diagnosis SAP Note:
files usually large files can indicate a the SAP HANA studio for de­ Files 1977208
problem with the database. tails.

52 Crashdump files Identifies new crashdump files that Check the contents of the Diagnosis SAP Note:
have been generated in the trace dump files. Files 1977218
directory of the system.

53 Pagedump files Identifies new pagedump files that Check the contents of the Diagnosis SAP Note:
have been generated in the trace dump files. Files 1977242
directory of the system.

54 Savepoint dura­ Identifies long-running savepoint Check disk I/O performance. Backup CPU Related
tion operations. Root Causes
and Solutions,
I/O Related
Root Causes
and Solutions,
SAP Note:
1977220

55 Columnstore Determines how many columns in Check sizing with respect to Memory Memory Prob­
unloads columnstore tables have been un­ data distribution. lems, SAP Note:
loaded from memory. This can indi­ 1977207
cate performance issues.

56 Python trace ac­ Determines whether or not the py­ If no longer required, deacti­ Diagnosis SAP Note:
tivity thon trace is active and for how vate the python trace in the Files 1977098
long. The python trace affects sys­ relevant configuration file.
tem performance.

57 Instance secure Determines if the instance secure Check and make sure that Security SAP Note:
store file system store in the file system (SSFS) of the instance SSFS is accessi­ 1977221
(SSFS) inacces­ your SAP HANA system is accessi­ ble to the database.
sible ble to the database.

58 Plan cache size Determines whether or not the plan Currently Alert 58 is inactive Memory SAP Note:
cache is too small. and replaced by Alert 91. 1977253
Please activate Alert 91 -
Plan Cache Hit Ratio

59 Percentage of Determines the percentage of Investigate blocking and Sessions/ Transactional

transactions transactions that are blocked. blocked transactions and if Transac­ Problems, SAP
blocked appropriate cancel some of tions Note: 2081856
them.

SAP HANA Troubleshooting and Performance Analysis Guide

Reference: Alerts PUBLIC 193
ID Name Description User Action Category Further Infor­
mation

60 Sync/Async Identifies a bad trigger asynchro­ Please refer to SAP note Disk I/O Related
read ratio nous read ratio. This means that 1930979. Root Causes
asynchronous reads are blocking and Solutions,
and behave almost like synchro­ SAP Note:
nous reads. This might have nega­ 1965379
tive impact on SAP HANA I/O per­
formance in certain scenarios.

61 Sync/Async Identifies a bad trigger asynchro­ Please refer to SAP note Disk I/O Related
write ratio nous write ratio. This means that 1930979. Root Causes
asynchronous writes are blocking and Solutions,
and behave almost like synchro­ SAP Note:
nous writes. This might have nega­ 1965379
tive impact on SAP HANA I/O per­
formance in certain scenarios.

62 Expiration of da­ Identifies database users whose Change the password of the Security SAP Note:
tabase user password is due to expire in line database user. 2082406
passwords with the configured password pol­
icy. If the password expires, the
user will be locked. If the user in
question is a technical user, this
may impact application availability.
It is recommended that you disable
the password lifetime check of
technical users so that their pass­
word never expires (ALTER USER
<username> DISABLE PASSWORD
LIFETIME).

63 Granting of Determines if the internal support Check if the corresponding Security SAP Note:
SAP_INTER­ role (SAP_INTERNAL_HANA_SUP­ users still need the role. If 2081857
NAL_HANA_SU PORT) is currently granted to any not, revoke the role from
PPORT role database users. them.

64 Total memory Determines what percentage of the Consider exporting the con­ Memory SAP Note:
usage of table- effective memory allocation limit is tent of the table and then 2081869
based audit log being consumed by the database truncating the table.
table used for table-based audit
logging. If this table grows too
large, the availability of the data­
base could be impacted.

65 Runtime of the Determines whether or not the Investigate why the log Backup SAP HANA Ad­
log backups cur­ most recent log backup terminates backup runs for too long, and ministration
rently running in the given time. resolve the issue. Guide, SAP
Note: 2081845

66 Storage snap­ Determines whether or not the pe­ Investigate why the storage Backup SAP HANA Ad­
shot is prepared riod, during which the database is snapshot was not confirmed ministration
prepared for a storage snapshot, or abandoned, and resolve Guide, SAP
exceeds a given threshold. the issue. Note: 2081405

67 Table growth of Determines the growth rate of row­ Try to reduce the size of row Memory SAP Note:
rowstore tables store tables store table by removing un­ 2054411
used data

SAP HANA Troubleshooting and Performance Analysis Guide

194 PUBLIC Reference: Alerts
 ID Name Description User Action Category Further Infor­
mation

68 Total memory Determines the current memory Investigate memory usage Memory SAP Note:
usage of row size of a row store used by a serv­ by row store tables and con­ 2050579
store ice sider cleanup of unused data

69 Enablement of Determines whether automatic log Enable automatic log Backup SAP HANA Ad­
automatic log backup is enabled. backup. For more details ministration
backup please see SAP HANA Ad­ Guide, SAP
ministration Guide. Note: 2081360

70 Consistency of Verifies the consistency of sche­ Contact SAP support. Availabil­

internal system mas and tables in internal system ity
components af­ components (for example, the re­
ter system up­ pository) after a system upgrade.
grade

71 Row store frag­ Check for fragmentation of row Implement SAP Note Memory SAP Note
mentation store. 1813245. 1813245

72 Number of log Determines the number of log seg­ Make sure that log backups Backup
segments ments in the log volume of each are being automatically cre­
serviceCheck for number of log ated and that there is
segments. enough space available for
them. Check whether the
system has been frequently
and unusually restarting
services. If it has, then re­
solve the root cause of this
issue and create log backups
as soon as possible.

73 Overflow of row­ Determines the overflow ratio of Identify the connection or Memory Transactional
store version the rowstore version space. transaction that is blocking Problems
space version garbage collection.
You can do this in the SAP
HANA studio by executing
"MVCC Blocker Statement"
and "MVCC Blocker Transac­
tion" available on the System
Information tab of the Ad­
ministration editor. If possi­
ble, kill the blocking connec­
tion or cancel the blocking
transaction.

SAP HANA Troubleshooting and Performance Analysis Guide

Reference: Alerts PUBLIC 195
ID Name Description User Action Category Further Infor­
mation

74 Overflow of met­ Determines the overflow ratio of Identify the connection or Memory Transactional
adata version the metadata version space. transaction that is blocking Problems
space version garbage collection.
You can do this in the SAP
HANA studio by executing
"MVCC Blocker Statement"
and "MVCC Blocker Transac­
tion" available on the System
Information tab of the Ad­
ministration editor. If possi­
ble, kill the blocking connec­
tion or cancel the blocking
transaction.

75 Rowstore ver­ Determines whether the rowstore Identify the connection or Memory Transactional
sion space skew version chain is too long. transaction that is blocking Problems
version garbage collection.
You can do this in the SAP
HANA studio by executing
"MVCC Blocker Statement"
and "MVCC Blocker Transac­
tion" available on the System
Information tab of the Ad­
ministration editor. If possi­
ble, kill the blocking connec­
tion or cancel the blocking
transaction. For your infor­
mation, you can find table in­
formation by using query
"SELECT * FROM TABLES
WHERE TABLE_OID = <table
object ID>".

76 Discrepancy be­ Identifies discrepancies between Check operating system Configu­

tween host the server times of hosts in a scale- time settings. ration
server times out system.

77 Database disk Determines the total used disk Investigate the disk usage of Disk
usage space of the database. All data, the database. See system
logs, traces and backups are con­ view M_DISK_USAGE for
sidered. more details.

78 Connection be­ Identifies closed connections be­ Investigate why connections Availabil­ SAP HANA Ad­
tween systems tween the primary system and a are closed (for example, net­ ity ministration
in system repli­ secondary system. If connections work problem) and resolve Guide
cation setup are closed, the primary system is the issue.
no longer being replicated.

SAP HANA Troubleshooting and Performance Analysis Guide

196 PUBLIC Reference: Alerts
 ID Name Description User Action Category Further Infor­
mation

79 Configuration Identifies configuration parameters If the identified configuration Configu­ SAP HANA Ad­
consistency of that do not have the same value on parameter(s) should have ration ministration
systems in sys­ the primary system and a secon­ the same value in both sys­ Guide
tem replication dary system. Most configuration tems, adjust the configura­
setup parameters should have the same tion. If different values are
value on both systems because the acceptable, add the parame­
secondary system has to take over ter(s) as an exception in
in the event of a disaster. global.ini/[inifile_checker].

80 Availability of ta­ Monitors error messages related to Determine which tables en­ Availabil­
ble replication table replication. countered the table replica­ ity
tion error using system view
M_TABLE_REPLICAS, and
then check the correspond­
ing indexserver alert traces.

81 Cached view Determines how much memory is Increase the size of the Memory
size occupied by cached view cached view. In the "re­
sult_cache" section of the in­
dexserver.ini file, increase
the value of the "total_size"
parameter.

82 Timezone con­ Compares SAP HANA internal Update SAP HANA internal Configu­ SAP note
version timezone conversion with Operat­ timezone tables (refer to ration 1932132
ing System timezone conversion. SAP note 1932132).

83 Table consis­ Identifies the number of errors and Contact SAP support. Availabil­
tency affected tables detected by ity
_SYS_STATISTICS.Collec­
tor_Global_Table_Consistency.

84 Insecure in­ Determines whether the master Change the instance SSFS Security SAP HANA Ad­
stance SSFS en­ key of the instance secure store in master key as soon as possi­ ministration
cryption config­ the file system (SSFS) of your SAP ble. For more information, Guide
uration HANA system has been changed. If see the SAP HANA Adminis­
the SSFS master key is not tration Guide.
changed after installation, it cannot
be guaranteed that the initial key is
unique.

85 Insecure sys­ Determines whether the master Change the system PKI Security SAP HANA Ad­
temPKI SSFS key of the secure store in the file SSFS master key as soon as ministration
encryption con­ system (SSFS) of your system's in­ possible. For more informa­ Guide
figuration ternal public key infrastructure tion, see the SAP HANA Ad­
(system PKI) has been changed. If ministration Guide.
the SSFS master key is not
changed after installation, it cannot
be guaranteed that the initial key is
unique.

SAP HANA Troubleshooting and Performance Analysis Guide

Reference: Alerts PUBLIC 197
ID Name Description User Action Category Further Infor­
mation

86 Internal commu­ Determines whether the ports used The parameter [communica­ Security SAP Note
nication is con­ by SAP HANA for internal commu­ tion] listeninterface in 2183363
figured too nication are securely configured. If global.ini is not set to a se­
openly the "listeninterface" property in the cure value. Please refer to
"communication" section of the SAP Note 2183363 or the
global.ini file does not have the section on internal host
value ".local" for single-host sys­ name resolution in the SAP
tems and ".all" or ".global" for mul­ HANA Administration Guide.
tiple-host systems, internal com­
munication channels are externally
exploitable.

87 Granting of SAP Determines if support privileges for Check if the corresponding Security
HANA DI sup­ the SAP HANA Deployment Infra­ users still need the privi­
port privileges structure (DI) are currently granted leges. If not, revoke the privi­
to any database users or roles. leges from them.

88 Auto merge for Determines if the delta merge of a The delta merge was not Memory
column-store ta­ table was executed successfully or executed successfully for a
bles not. table. Check the error de­
scription in view
M_DELTA_MERGE_STATIS­
TICS and also Indexserver
trace.

89 Missing volume Determines if there is any volume Volume file missing, data­ Configu­
files file missing. base instance is broken, stop ration
immediately all operations
on this instance.

90 Status of HANA Determines if the system was not Install/update SAP HANA Configu­ SAP note
platform lifecy­ installed/updated with the SAP Database Lifecycle Manager ration 2078425
cle management HANA Database Lifecycle Manager (HDBLCM). Implement SAP
configuration (HDBLCM). note 2078425

91 Plan cache hit Determines whether or not the plan Increase the size of the plan Memory
ratio cache hit ratio is too low. cache. In the "sql" section of
the indexserver.ini file, in­
crease the value of the
"plan_cache_size" parame­
ter.

92 Root keys of Not al services that persist data Trigger a savepoint for this Security
persistent serv­ could be reached the last time the service or flush the SSFS
ices are not root key change of the data volume cache using hdbcons
properly encryption service was changed.
synchronized As a result, at least one service is
running with an old root key.

93 Streaming Li­ Determines how many days until Obtain a valid license and in­ Availabil­
cense expiry your streaming license expires. stall it. For the exact expira­ ity
Once your license expires, you can tion date, see the monitoring
no longer start streaming projects. view M_LICENSES.

SAP HANA Troubleshooting and Performance Analysis Guide

198 PUBLIC Reference: Alerts
 ID Name Description User Action Category Further Infor­
mation

94 Log replay back­ System Replication secondary site Investigate on secondary Availabil­
log for system has a higher log replay backlog site, why log replay backlog ity
replication sec­ than expected. is increased
ondary

95 Availability of Determine the Data Quality refer­ Download the latest Data Availabil­ SAP HANA Ad­
Data Quality ref­ ence data expiration dates. Quality reference data files ity ministration
erence data (di­ and update the system. (For Guide
rectory files) more details about updating
the directories, see the Ad­
ministration Guide.)

96 Long-running Identifies all long-running tasks. Investigate the long-running 0

tasks tasks. For more information,
see the task statistics tables
or views in _SYS_TASK
schema and trace log.

97 Granting of SAP Determines if the container import Check if the identified users Security
HANA DI con­ feature of the SAP HANA Deploy­ still need the privileges. If
tainer import ment Infrastructure (DI) is enabled not, revoke the privileges
privileges and if import privileges for SAP from them and disable the
HANA DI containers are currently SAP HANA DI container im­
granted to any database users or port feature.
roles.

98 LOB garbage Determines whether or not the lob Activate the LOB garbage Configu­
collection activ­ garbage collection is activated. collection using the corre­ ration
ity sponding configuration pa­
rameters.

99 Maintenance Checks the installed SP version Please consider upgrading to Configu­

Status against the recommended SP ver­ the recommended SP ver­ ration
sion. sion.

500 Dbspace usage Checks for the dbspace size usage. Investigate the usage of Disk
dbspace and increase the
size.

501 Dbspace status Determines whether or not all Investigate why the dbspace Availabil­
dbspaces are available. is not available. ity

502 Dbspace file sta­ Determines whether or not all Investigate why the dbspace Availabil­
tus dbspace files are available. file is not available. ity

503 Inconsistent Configuration between HANA and Investigate the alert and con­ Security
Persistence En­ HANA Extended Storage do not figure the persistence En­
cryption match. Data may not be com­ cryption as required.
pletely protected.

600 Inactive Stream­ Identifies inactive Streaming appli­ Investigate why the Stream­ Availabil­
ing applications cations. ing application is inactive, for ity
example, by checking the
Streaming application's
trace files.

SAP HANA Troubleshooting and Performance Analysis Guide

Reference: Alerts PUBLIC 199
ID Name Description User Action Category Further Infor­
mation

601 Inactive Stream­ Identifies inactive Streaming Investigate why the Stream­ Availabil­
ing project man­ project managed adapters. ing project managed adapter ity
aged adapters is inactive, for example, by
checking the trace files.

602 Streaming Determines what percentage of to­ Investigate memory usage of Memory
project physical tal physical memory available on the streaming project.
memory usage the host is used for the streaming
project.

603 Streaming Determines the percentage CPU Investigate CPU usage. CPU
project CPU us­ usage for a streaming project on
age the host and therefore whether or
not CPU resources are running out.

604 Number of pub­ Identify the large publishers of Investigate whether these Configu­
lishers of streaming project. Make sure that publishers are created inten­ ration
streaming they will not break the streaming tionally.
project project.

605 Number of sub­ Identify the large subscribers of Investigate whether these Configu­
scribers of streaming project. Make sure that subscribers are created in­ ration
streaming they will not break the streaming tentionally.
project project.

606 Row throughput Identify which subscriber of Investigate why the sub­ Configu­
of subscriber of streaming project has low through­ scriber works slowly. ration
streaming put measured in rows per second.
project

607 Transaction Identify which subscriber of Investigate why the sub­ Configu­
throughput of streaming project has transaction scriber works slowly. ration
subscriber of throughput measured in transac­
streaming tions per second.
project

608 Row throughput Identify which publisher of stream­ Investigate why the publisher Configu­
of publisher of ing project has low throughput works slowly. ration
streaming measured in rows per second.
project

609 Transaction Identify which publisher of stream­ Investigate why the publisher Configu­
throughput of ing project has transaction works slowly. ration
publisher of throughput measured in transac­
streaming tions per second.
project

610 Bad rows of Identify which project managed Investigate why the adapter Configu­
project man­ adapter has much rows with error. has such much rows with er­ ration
aged adapter ror.

611 High latency of Identify which project managed Investigate why the adapter Configu­
project man­ adapter has high latency. has high latency. ration
aged adapter

SAP HANA Troubleshooting and Performance Analysis Guide

200 PUBLIC Reference: Alerts
 ID Name Description User Action Category Further Infor­
mation

612 Large queue of Identify which stream of streaming Investigate why the stream Configu­
stream of project has large queue. has large queue. ration
streaming
project

613 Large store of Identify which stream of streaming Investigate why the stream Configu­
stream of project has large store. has large store. ration
streaming
project

700 Agent availabil­ Determines how many minutes the Investigate connection of Availabil­
ity agent has been inactive. agent and check if agent is ity
up and running.

701 Agent memory Determines what percentage of to­ Investigate which adapter or Memory
usage tal available memory on agent is processes use a lot of mem­
used. ory.

710 Remote Sub­ Checks for recent exceptions in re­ Investigate the error mes­ 0
scription excep­ mote subscriptions and remote sage and the error code and
tion sources. restart the remote subscrip­
tion if necessary.

SAP HANA Troubleshooting and Performance Analysis Guide

Reference: Alerts PUBLIC 201
6 HANA Options: Performance Analysis

SAP HANA options provide additional functions you can use together with the SAP HANA base edition. These
topics describe what steps you can take to identify and resolve specific performance issues and what you can
do to enhance the performance of your SAP HANA options.

Caution
SAP HANA server software and tools can be used for several SAP HANA platform and options scenarios as
well as the respective capabilities used in these scenarios. The availability of these is based on the available
SAP HANA licenses and the SAP HANA landscape, including the type and version of the back-end systems
the SAP HANA administration and development tools are connected to. There are several types of licenses
available for SAP HANA. Depending on your SAP HANA installation license type, some of the features and
tools described in the SAP HANA platform documentation may only be available in the SAP HANA options
and capabilities, which may be released independently of an SAP HANA Platform Support Package Stack
(SPS). Although various features included in SAP HANA options and capabilities are cited in the SAP HANA
platform documentation, each SAP HANA edition governs the options and capabilities available. Based on
this, customers do not necessarily have the right to use features included in SAP HANA options and
capabilities. For customers to whom these license restrictions apply, the use of features included in SAP
HANA options and capabilities in a production system requires purchasing the corresponding software
license(s) from SAP. The documentation for the SAP HANA optional components is available in SAP Help
Portal at http://help.sap.com/hana_options. If you have additional questions about what your particular
license provides, or wish to discuss licensing features available in SAP HANA options, please contact your
SAP account team representative.

6.1 SAP HANA Dynamic Tiering

Identify and resolve specific performance issues and enhance the performance of SAP HANA dynamic tiering.

Note
Troubleshooting information for SAP HANA dynamic tiering is in the SAP HANA Dynamic Tiering:
Administration Guide.

6.1.1 Tools and Tracing

This section gives you an overview of the tools and tracing options available for SAP HANA dynamic tiering.

SAP HANA Troubleshooting and Performance Analysis Guide

202 PUBLIC HANA Options: Performance Analysis
6.1.1.1 Federation Trace
Federation trace can be turned ON to diagnose most issues with SAP HANA dynamic tiering.

Federation trace generates tracing information in the indexserver.ini trace file.

To enable federation trace:

ALTER SYSTEM ALTER CONFIGURATION ('indexserver.ini', 'SYSTEM') SET ('trace',

'fedtrace') = 'debug' WITH RECONFIGURE;
ALTER SYSTEM ALTER CONFIGURATION ('indexserver.ini', 'SYSTEM') SET ('trace',
'federationexecution') = 'debug' WITH RECONFIGURE;

6.1.2 Query Plan Analysis

The query plan shows the various operators involved in the query execution.

Queries referencing both SAP HANA tables and SAP HANA dynamic tiering tables are either:

● Executed in SAP HANA by pulling data from SAP HANA dynamic tiering
● Relocated to SAP HANA dynamic tiering, where the data is pulled from SAP HANA

Generally, since SAP HANA dynamic tiering involves tables with large amounts of data, it may be preferable to
use the latter strategy. Another reason to use the latter strategy is when the SAP HANA dynamic tiering table
is too large to fit in SAP HANA.

If your query involves both SAP HANA tables and SAP HANA dynamic tiering tables and you are experiencing
poor performance, you should review the query plan. Review the visual query plan that shows the timings for
various sub-trees. Alternatively, you can query M_REMOTE_STATEMENTS to show timing results for query
fragments executed on SAP HANA dynamic tiering. If the timing shown is small, and you think the optimizer is
shipping the upper limit of query fragments, then SAP HANA dynamic tiering is probably not the cause of the
performance problem.

The Remote Row Scan operator deals with a query fragment executed by an SAP HANA dynamic tiering
node. If this operator is directly under the top-level Project node, then the entire query has been either
pushed down or relocated to an SAP HANA dynamic tiering node. Generally, relocating the query yields better
performance.

Changing the Execution Strategy to Remote

If the execution strategy 'auto' mode is not yielding the best plan, try the following procedure to change the
execution strategy from 'auto' to 'remote':

1. Clear the query plan cache.

ALTER SYSTEM CLEAR SQL PLAN CACHE;

2. Change the execution strategy from 'auto' to 'remote':

ALTER SYSTEM ALTER CONFIGURATION ('esserver.ini', 'SYSTEM') SET

('row_engine', 'execution_strategies') = 'remote' WITH RECONFIGURE;

SAP HANA Troubleshooting and Performance Analysis Guide

HANA Options: Performance Analysis PUBLIC 203
Changing the Execution Strategy to Auto

If the execution strategy is set to 'auto', then the optimizer chooses the best strategy for executing the query:
either relocating the query to SAP HANA dynamic tiering, or executing the query in SAP HANA. In most cases,
'auto' provides best performance.

1. Change the execution strategy from 'remote' to 'auto':

ALTER SYSTEM ALTER CONFIGURATION ('esserver.ini', 'SYSTEM') SET

('row_engine', 'execution_strategies') = 'auto' WITH RECONFIGURE;

If neither execution strategy improves performance, there may be a capability issue. The query optimizer
decides what to push down to SAP HANA dynamic tiering based on the capability supported by the option. If
the query deals with some operator, builtin, or other item that SAP HANA dynamic tiering does not
understand, then it lets the SAP HANA execution engine compensate for it. Review your query to see if there
are any unsupported operators, or builtins and see if you can rewrite the query without them. Finally, you may
be missing statistics on the tables that may prevent the optimizer from choosing an optimal query plan.

Related Information

Statistics [page 204]

6.1.2.1 Statistics

Statistics help the query optimizer in choosing the right query plan. Missing statistics may prevent the query
optimizer from selecting the optimal query plan.

When no statistics for an extended table are present, the query optimizer assumes the table size to be 1 million
rows. If the actual table has significantly different number of rows, then the query plan chosen may not be
optimal. To ensure that optimizer has the correct information, we recommend that you create statistics on the
extended tables.

SAP HANA currently supports two types of statistics:

Histogram Creates a multi-bucket histogram. Histogram statistics provides much better estimates for
range queries.

Simple Creates simple statistics for each column in the table. Statistics are: minimum value, maximum
value, total count of values, distinct count of values, and null counts.

We recommend that simple statistics, at the very least, are present on key columns of extended tables.

To create simple statistics on column c1:

CREATE STATISTICS ON TAB1 (c1) TYPE SIMPLE

To create simple statistics on all columns:

CREATE STATISTICS ON TAB1 TYPE SIMPLE

SAP HANA Troubleshooting and Performance Analysis Guide

204 PUBLIC HANA Options: Performance Analysis
To create histogram-type statistics on column c1:

CREATE STATISTICS ON TAB1 (c1) TYPE HISTOGRAM

6.1.3 Data Loading Performance

SAP HANA dynamic tiering supports all data loading methods for extended tables. This section explores the
various mechanisms for inserting data into extended tables, and recommends the optimal loading
mechanisms.

In general, SAP HANA dynamic tiering is optimized for batch writes, like SAP HANA. Singleton writes are not
the best use case for SAP HANA dynamic tiering, although singleton writes are supported.

6.1.3.1 IMPORT FROM Statement

If you have a CSV file for the data to be loaded, the IMPORT FROM statement is by far the best method for
loading data into an SAP HANA dynamic tiering table.

The IMPORT FROM statement is sent directly to the SAP HANA dynamic tiering node for the data load.
Currently, IMPORT FROM does not support the THREADS parameter on the import command and hence it is
better to break the files into multiple files if you are planning on loading hundreds of millions of rows.

When dealing with a delta-enabled extended table, you can run these IMPORT FROM statements in parallel if
required. However, note that this will put heavy demand on the delta memory. Make sure delta memory is
appropriately configured. Alternatively, you can do this load serially.

When importing very large amounts of data into an extended table, use multiple files in the same IMPORT
FROM statement. Breaking into multiple files for IMPORT FROM yields better performance than a single file
import.

6.1.3.2 INSERT Statement with SELECT Statement

If the data is present in another SAP HANA table, then INSERT-SELECT is a better loading method than the
IMPORT FROM statement.

The query optimizer tries to relocate the INSERT-SELECT on an extended table to the SAP HANA dynamic
tiering node. The SAP HANA dynamic tiering node does a parallel fetch from SAP HANA, thereby speeding up
the INSERT-SELECT.

SAP HANA Troubleshooting and Performance Analysis Guide

HANA Options: Performance Analysis PUBLIC 205
6.1.3.3 Parameterized Array Inserts

Array insert is by far the most optimal mechanism to load data into an extended table.

SAP HANA dynamic tiering converts an array-insert into a LOAD statement on the SAP HANA dynamic tiering
node.

Bulk load is controlled by the bulk_inserts_as_load and bulk_load_as_binary parameters. Both

parameters are 'true' by default.

If you need to re-enable the defaults:

1. Re-enable the bulk load mechanism for optimizing array inserts:

ALTER SYSTEM ALTER CONFIGURATION ('esserver.ini', 'SYSTEM') SET

('row_engine', 'bulk_inserts_as_load') = 'true' WITH RECONFIGURE;

2. Re-enable binary load (instead of the ASCII load):

ALTER SYSTEM ALTER CONFIGURATION ('esserver.ini', 'SYSTEM') SET

('row_engine', 'bulk_load_as_binary') = 'true' WITH RECONFIGURE;

SAP HANA Troubleshooting and Performance Analysis Guide

206 PUBLIC HANA Options: Performance Analysis
Important Disclaimer for Features in SAP
HANA Platform, Options and Capabilities

SAP HANA server software and tools can be used for several SAP HANA platform and options scenarios as
well as the respective capabilities used in these scenarios. The availability of these is based on the available
SAP HANA licenses and the SAP HANA landscape, including the type and version of the back-end systems the
SAP HANA administration and development tools are connected to. There are several types of licenses
available for SAP HANA. Depending on your SAP HANA installation license type, some of the features and
tools described in the SAP HANA platform documentation may only be available in the SAP HANA options and
capabilities, which may be released independently of an SAP HANA Platform Support Package Stack (SPS).
Although various features included in SAP HANA options and capabilities are cited in the SAP HANA platform
documentation, each SAP HANA edition governs the options and capabilities available. Based on this,
customers do not necessarily have the right to use features included in SAP HANA options and capabilities.
For customers to whom these license restrictions apply, the use of features included in SAP HANA options and
capabilities in a production system requires purchasing the corresponding software license(s) from SAP. The
documentation for the SAP HANA optional components is available in SAP Help Portal at http://
help.sap.com/hana_options. If you have additional questions about what your particular license provides, or
wish to discuss licensing features available in SAP HANA options, please contact your SAP account team
representative.

SAP HANA Troubleshooting and Performance Analysis Guide

Important Disclaimer for Features in SAP HANA Platform, Options and Capabilities PUBLIC 207
Important Disclaimers and Legal Information

Coding Samples
Any software coding and/or code lines / strings ("Code") included in this documentation are only examples and are not intended to be used in a productive system
environment. The Code is only intended to better explain and visualize the syntax and phrasing rules of certain coding. SAP does not warrant the correctness and
completeness of the Code given herein, and SAP shall not be liable for errors or damages caused by the usage of the Code, unless damages were caused by SAP
intentionally or by SAP's gross negligence.

Accessibility
The information contained in the SAP documentation represents SAP's current view of accessibility criteria as of the date of publication; it is in no way intended to be
a binding guideline on how to ensure accessibility of software products. SAP in particular disclaims any liability in relation to this document. This disclaimer, however,
does not apply in cases of willful misconduct or gross negligence of SAP. Furthermore, this document does not result in any direct or indirect contractual obligations
of SAP.

Gender-Neutral Language
As far as possible, SAP documentation is gender neutral. Depending on the context, the reader is addressed directly with "you", or a gender-neutral noun (such as
"sales person" or "working days") is used. If when referring to members of both sexes, however, the third-person singular cannot be avoided or a gender-neutral noun
does not exist, SAP reserves the right to use the masculine form of the noun and pronoun. This is to ensure that the documentation remains comprehensible.

Internet Hyperlinks
The SAP documentation may contain hyperlinks to the Internet. These hyperlinks are intended to serve as a hint about where to find related information. SAP does
not warrant the availability and correctness of this related information or the ability of this information to serve a particular purpose. SAP shall not be liable for any
damages caused by the use of related information unless damages have been caused by SAP's gross negligence or willful misconduct. All links are categorized for
transparency (see: http://help.sap.com/disclaimer).

SAP HANA Troubleshooting and Performance Analysis Guide

208 PUBLIC Important Disclaimers and Legal Information
SAP HANA Troubleshooting and Performance Analysis Guide
Important Disclaimers and Legal Information PUBLIC 209
go.sap.com/registration/
contact.html

© 2016 SAP SE or an SAP affiliate company. All rights reserved.

No part of this publication may be reproduced or transmitted in any
form or for any purpose without the express permission of SAP SE
or an SAP affiliate company. The information contained herein may
be changed without prior notice.
Some software products marketed by SAP SE and its distributors
contain proprietary software components of other software
vendors. National product specifications may vary.
These materials are provided by SAP SE or an SAP affiliate company
for informational purposes only, without representation or warranty
of any kind, and SAP or its affiliated companies shall not be liable for
errors or omissions with respect to the materials. The only
warranties for SAP or SAP affiliate company products and services
are those that are set forth in the express warranty statements
accompanying such products and services, if any. Nothing herein
should be construed as constituting an additional warranty.
SAP and other SAP products and services mentioned herein as well
as their respective logos are trademarks or registered trademarks
of SAP SE (or an SAP affiliate company) in Germany and other
countries. All other product and service names mentioned are the
trademarks of their respective companies.
Please see http://www.sap.com/corporate-en/legal/copyright/
index.epx for additional trademark information and notices.