Tips & Tricks for Using LVM

Effectively / Intro to VxVM

Renay Gaye
Hewlett-Packard
[email protected]

LVM Concepts

LVM Concepts-Physical Volume

LVM Managed Disk
Each PV is assigned a unique PVID

pvid

pvcreate /dev/rdsk/cntndn

LVM Concepts-Volume Group

One or more physical volumes
pool of physical extents
mkdir /dev/vgname
pvra
vgra
extents

pvid

vgid

pvid

vgid

mknod /dev/vgname/group c 64
0xnn0000
vgcreate vgname /dev/dsk/cntndn ...

Volume Group Map

vgid

pvid

vgid

pvid

map

map

pv 0
pv 1
pv 2

extent size (-s n)

* max extents/pv (-e n)
______________________
max. useable space/disk

~
~
pv 15

pe pe
2 3

~
~

pe pe
0 1

pe
1016

LVM Device Files

/dev

dsk

rdsk

vg01

c0t5d0

c0t5d0

c0t6d0

c0t6d0

group
lvol1

V G device
file
Block LV
device files

lvol2
Block PV
device files

Raw PV
device files

rlvol1
rlvol2

Raw LV
device files

/etc/lvmtab
Non-ascii file
Records volume
group/physical volume
associations
Used by many LVM
commands for sanity
checking

example:

/dev/vg00 (vgid)
/dev/dsk/c0t6d0 (vgid,pvid)
/dev/vg01 (vgid)
/dev/dsk/c1t2d1 (vgid,pvid)
/dev/dsk/c1t2d2 (vgid,pvid)

Volume Group Activation

Done automatically when vg is created
automatically at boot (/etc/lvmrc)
required in order to access any lvols
00
01
pvid

vgid

vgra

pvid

vgid

vgra

vgchange -a y vgname
vgchange -a r vgname

Memory

LVM Concepts-Quorum
More than 1/2 disks in a vg
required to activate the
volume group
Can override:

vg01

vgchange -a y -q n vg01
pvid

vgid

pvid

vgid

Booting without quorum:

ISL> hpux -lq

LVM Concepts-Logical Volume

vgid

pvid
map

vgid

pvid
map

pv 0 01
pv 1
pv 2

lvcreate -L n vgname
lvcreate -L n -C y vgname
lvcreate -l n vgname

~
~
pv 15

lvcreate -L n -n name vgname

01

pe pe pe
2 3 4
01

01

01

~
~

pe pe
0 1

pe
1016

Placing logical volumes

vgid

pvid

map

pv 0 01

01

01

01

01

pv 1 02

02

02

02

02

pv 2

lvcreate -n datalv vgdata

lvextend -L 500
/dev/vgdata/datalv
/dev/dsk/c4t2d0

~
~
pv 15

~
~

map

vgid

pvid

pe pe pe
2 3 4

~
~

pe pe
0 1

02

pe
1016

Multi-pathing solutions

Multipathing Solutions with LVM

LVM pvlinks
active/passive
no additional software required

Auto path
active/active, load balancing
supported for XP and VA arrays
additional cost product

Powerpath
active/active, load balancing
supported for EMC arrays
additional cost product

PV links
pvcreate /dev/rdsk/c5t5d0
pvcreate /dev/rdsk/c5t5d1

pvid

vgid

pvid

vgid

mkdir /dev/vg01
mknod /dev/vg01/group c 64
0x010000
vgcreate vg01
/dev/dsk/c5t5d0 /dev/dsk/c7t5d0
/dev/dsk/c7t5d1 /dev/dsk/c5t5d1

c5t5d0 (p)
c7t5d0 (a)

c7t5d1 (p)
c5t5d1 (a)

PV links-switching the order

Temporary Change
pvchange -s /dev/dsk/c7t5d0

pvid

vgid

pvid

vgid

Permanent Change
vgreduce vg01 /dev/dsk/c5t5d0
vgextend vg01 /dev/dsk/c5t5d0

c5t5d0 (p) (a)

c7t5d0 (a) (p)

c7t5d1 (p)
c5t5d1 (a)

Auto Path-Dynamic load

balancing
Load balancing
functionality
Supports up to 8 paths from a server
to an end device
Provides dynamic load balancing
across all paths to an end device
Choose from 4 load balancing
policies, including no load balancing
Supports the XP and VA disk arrays
Load balancing supported in
clustered environments

Server
Auto Path
HBA HBA HBA HBA HBA HBA HBA HBA

FC FC FC FC FC FC FC FC
Port Port Port Port Port Port Port Port
C HIP Board 1
C HIP Board 2

Array

LDEV

Auto Path driver

Application

A pseudo driver
The Auto Path driver is
layered between the LVM
(Logical Volume
Manager) and the SCSI
device driver
The driver provides the
Command Line Interface

File System
Logical Volum e Manager
Auto Path
Drivers
(SCSI device and FC interface)

FC HBA

FC HBA

Disk Array

Moving Data

Moving Disks
Three Step Process
1. Remove definition of volume group
2. Move disk(s)
3. Add definition ofvolume group
Two com mands
vgexport(1m)
vgimport(1m)

Volume Group Definition

Review
/etc/lvmtab

name of Logical Volumes

/dev/vg01

/dev/vg01
group

/dev/dsk/c0t5d0

64 0x010000

lvol1 64 0x010001
rlvol1 64 0x010001
databaselv 64 0x010002
rdatabaselv 64 0x010002
V GR A + PVRA
-PV ID
- VG ID
PE to LE map

Exporting a Volume Group

Syntax:
vgexport [-p][-v][-m file]VG
-p
Preview actions only
-v
verbose
-m
used to specify a map fileforlogical volume names
Removes volume group definition from the system completely by updating
/etc/lvmtab and kernel memory.
The volume group must first be deactivated with vgchange(1m).
Example:
vgchange -a n /dev/vg01
vgexport -v -m /etc/lvmconf/vg01.map

/dev/vg01

Importing a Volume Group

Syntax:
vgimport [-p][-v][-m file] VG PV [PV...]

-p Preview actions only

-v verbose
-m used to specify a map file for logical volume names

Example:
mkdir /dev/vg01
mknod /dev/vg01/group c 64 0x010000
vgimport -v -m /etc/lvmconf/vg01.map /dev/vg01 /dev/dsk/c0t1d0
vgchange -a y /dev/vg01
vgcfgbackup vg01

Moving LVM Data

Volume group

pvmove

free

free
c0t5d0

c0t4d0

Syntax:
pvmove [-n lv] from_PV [to_PV]

Example:
pvmove -n /dev/vg01/lvol1 /dev/dsk/c0t5d0 /dev/dsk/c0t4d0

Renaming LVM objects

Renaming Logical Volumes

lvol names are not stored in the LVM
maps or in /etc/lvmtab
Simply rename device files and update
/etc/fstab if necessary

Renaming Volume Groups

Volume group name is kept in
/etc/lvmtab and is used as the directory
name to anchor the device files for the
group
Use vgexport and vgimport to rename
volume group

LVM Mirroring

Mirrored Volumes
In a mirrored logical volume ...
Each logical volume consists of one or more Logical Extents(LEs).
Each LE maps to two orthree Physical Extents(PEs) on disk.
The logical volume remains accessible even when one ofthe disksis unavailable.
disk1

disk2

PE0

lvol1
LE0

PE0

PE1

LE1

PE1

PE2

LE2

PE2

PE3

PE3

PE4

lvol2
LE0

PE4

PE5

LE1

PE5

Extending and Reducing

Mirrors
Create a new, mirrored logical volume:
# lvcreate -m 1 -L 16 -n myfs1 vg01
Mirror an existing logicalvolume:
# lvextend -m 1 /dev/vg01/myfs1
Mirror an existing logicalvolume to a specific disk:
# lvextend -m 1 /dev/vg01/myfs1 /dev/dsk/c0t3d0
Add a second mirror:
# lvextend -m 2 /dev/vg01/myfs1
Remove a logicalvolumes mirrors:
# lvreduce -m 0 /dev/vg01/myfs1 /dev/dsk/c0t3d0
Check a mirrored logical volumes status:
# lvdisplay -v /dev/vg01/myfs1

Mirrored I/O Scheduling

Parallel(-d p)

Sequential(-d s)
Read in PV order

Read

Access PV with lowest

outstanding I/Os

Write

Schedule writes
simultaneously to all PVs

Schedule writes in
PV order

MWC/MCR
Writes are recorded in MWC in memory
MCR record written to disk when a write is done to
a logical track group not already recorded
Write IO

M WC
LTG Status
0

After a crash only "dirty" LTGs need be resynced

1
2
LTG5

PVRA
V GR A

Mirror Consistency Recovery Options

Advantage

Disadvantage
Runtime overhead to
write MCR record

M W C/ M CR
-M y

Fastrecovery on
a system crash

NO M W C
-M n -c y

No extra
overhead at

Slow recovery on
a system crash

runtime
3

N O NE
-M n -c n

Application can
do necessary
recovery

No system managed
recovery

LVM Boot Disks

LVM Boot volume structure

LIF Directory
PVRA
BDR A

ISL,HPUX,AUTO,LABEL

Hardware path/device file of disks

in volume group

LIF

boot, root, dump, swap lvols

V GR A

Maintained with the lvlnboot

command

PE

Mirroring the LVM Boot Disk

Initialize disk for LVM. Leave room for BDRA/LIF

pvcreate -B /dev/rdsk/cntndn

Add disk to root VG

vgextend vg00 /dev/rdsk/cntndn

Install boot files in LIF area

mkboot /dev/rdsk/cntndn

Change the auto file on both the primary and alternate boot disk
mkboot -a "hpux -lq" /dev/rdsk/cntndn

Mirror each of the lvols in the root vg

lvextend -m 1 /dev/vg00/lvoln /dev/dsk/cntndn

Add the mirror disk definition to /stand/bootconf

Booting When BDRA is

damaged
Boot system into maintenance mode
ISL> hpux -lm
Activate vg00
vgchange -a y vg00
Use lvlnbootto examine/repair BDRA
lvlnboot -v
lvlnboot -b /dev/vg00/lvol1
lvlnboot -r /dev/vg00/lvol3
lvlnboot -s /dev/vg00/lvol2
lvlnbtoo -d /dev/vg00/lvol2
Reboot the system
reboot

Moving the Boot Disk

Problem: /etc/lvmtab contains the old
device files for the boot disk
Solution: boot into maintenace mode,
export and reimport volume group

Problem: The BDRA and Label files

contain the old device information
Solution: use lvlnboot to fix

Cookbook for moving root disk

Boot from new device. Reply Y to interact with IPL
ISL> hpux -lm
# vgexport -v -m vg00.map vg00
# mkdir /dev/vg00
# mknod /dev/vg00/group c 64 0x000000
# vgimport -v -m vg00.map vg00 /dev/dsk/new_device_file
# vgchange -a y vg00
# vgcfgbackup vg00
# lvlnboot -R
# lvlnboot -v

Recovering Corrupt LVM

Information

Valuable Data?
The variouslvm data structures are crucialtothe continued availability of
logicalvolumes
/etc/lvmtab
Sanity check
vgdisplay -v

Kernel memory

Active
00
V Gs

01 02 03

/dev/vg01
/dev/dsk/c0t5d0
Cross
reference

PVRA/
V GR A

Backing up LVM Structures

vg01
/etc/lvmconf

LV M
Structures
File
System
File
System

vgcfgbackup

vg01.conf

vg00.conf

Done automatically by commands that modify

the configuration
Syntax:
vgcfgbackup [-f file]VG

Recovering LVM Structures

vg01
/etc/lvmconf

LV M
Structures

vgcfgrestore

vg01.conf

vg00.conf

Only restores the LV M structures.Contents of logical

volumes must be restored separately.

Syntax:
vgcfgrestore [-n VG|-f file][-o PV]PV
vgcfgrestore -l -n VG

/etc/lvmtab

LVM Control File

Not asciidata

vg00
/dev/dsk/c0t6d0
/dev/dsk/c0t5d0
/dev/dsk/c0t4d0
vg01
/dev/dsk.....
............

can use strings (1)to read theasciipart

Used primarily at boot up but also used to
sanity check com mands

Can be rebuiltiflost

PV ID

PV ID

PV ID

VG ID

VG ID

VG ID

All VG IDs the same - all disksin same vg

Recover or Repair
/etc/lvmtab
Syntax:
vgscan [-v][-p]
Semi-intelligentlook at every disk to classify it
importsthose that are known to be on this system
recommends the import of "new" disks
rebuilds /etc/lvmtab
Ways to confuse vgscan
redundant group files
"old" LV M structures

LVM Performance Tips

Use PVGs to Offload Busy

Controllers Volume Group vg01
PVG1

System

PVG2

Use Striping to Offload Busy

Drives
1 3 5
7 9 11

1 2 3
4 5 6
100%
Util

60%
Util

52 %
Util

60%
Util

System

System

2 4 6
8 10 12
5%
Util

50%
Util

Volume Group vg01

Without Striping

52 %
Util

50%
Util

Volume Group vg01

With Striping

LVM Distributed Allocation

Logical Volume
LE0
LE1
LE2
LE3
LE4
LE5
LE6
LE7
LE8
LE9
LE10
LE11

Volume Group vg01

LE0
LE4
LE8

LE1
LE5
LE9

LE2
LE6
LE10

LE3
LE7
LE11

PV1

PV2

PV3

PV4

PVG1

lvcreate -l 12 -s g -D y /dev/vg01

Mirroring and Striping

Logical Volume
LE0
LE1
LE2
LE3
LE4
LE5
LE6
LE7
LE8
LE9
LE10
LE11

Volume Group vg01

PVG1

LE0
LE4
LE8

PV1

PVG2

LE0
LE4
LE8

PV1

LE1
LE5
LE9

PV2
LE1
LE5
LE9

PV2

LE2
LE6
LE10

PV3
LE2
LE6
LE10

PV3

lvcreate -l 12 -m 1 -s g -D y /dev/vg01

LE3
LE7
LE11

PV4
LE3
LE7
LE11

PV4

LVM in an MC/ServiceGuard
Environment

MC/ServiceGuard
Configuration

System
A

Primary

Mirror

Volume Group

System
B

Step 1: Configure Volume

Group on First System

System A
Volume Group

Step 2: Import Volume Group

to Second System

System B

System A
Volume Group

VG Export and Import -s Option

vgexport -p -s -m <mapfilename> <vg_name>
create mapfile withoutremoving the VG,
and save the VGID forthe vgimportcommand
vgimport -s -m <mapfilename> <vg_name>
scan for disks that have the same VG ID asinthe mapfile

M ap File
VGID xxxxxxx
1 lvol_name1
2 lvol_name2
3 lvol_name3

System B

System A
vgexport -p -s -m /tmp/map vg01

Vol Grp

vgimport -s -m /tmp/map vg01

LVM Definitions-Both Nodes

Node 1

Node 2

/dev/vg01

/dev/vg01

group 64 0x010000

group 64 0x010000

lvol1 64 0x010001
rlvol1 64 0x010001

lvol1 64 0x010001
rlvol1 64 0x010001

databaselv 64 0x010002
rdatabaselv 64 0x010002

databaselv 64 0x010002
rdatabaselv 64 0x010002

/etc/lvmtab
/dev/vg01
/dev/dsk/c5t5d0
/dev/dsk/c7t5d0
/dev/dsk/c7t5d1
/dev/dsk/c5t5d1

pvid

vgid

pvid

vgid

/etc/lvmtab
/dev/vg01
/dev/dsk/c5t5d0
/dev/dsk/c7t5d0
/dev/dsk/c7t5d1
/dev/dsk/c5t5d1

Cluster Volume Group

vgchange -c y vg01

Node1

Node2

cluster id

cluster id

Volume Group Exclusive Activation

vgchange -a e vg01
Q. Do you
have the
volume group
activated?

A. No I
do not

Node1
cmlvmd

Node2
cmlvmd
cluster id

cluster id

Marking Volume Groups as

M C/ServiceGuard Volume Groups
Marking Volume Group
for MC/ServiceGuard

vgchange -c y VGName

Marking Volume Group

as non-M C/ServiceGuard

vgchange -c n VGName

Standard Volume Group

Activation

vgchange -a y VGName

Exclusive Volume Group

Activation

vgchange -a e VGName

Disk Array Data Replication

Issues

Problem-Duplicate vgid
vgid-12345
c1t3d1

vgid-12345
c1t3d2

vgprim(pvols)
vgid-12345
c2t6d1

vgid-12345
c2t6d2

vgsec(svols)
Disk Array

Solution-vgchgid
vgid-12345
c1t3d1

vgid-12345
c1t3d2

vgprim(pvols)

vgchgid /dev/rdsk/c2t6d1
/dev/rdsk/c2t6d2
mkdir /dev/vgsec
mknod /dev/vgsec/group c 64
0x020000
vgexport -pm mapfile vgprim

vgid-56789
c2t6d1

vgid-56789
c2t6d2

vgsec(svols)

vgimport -m mapfile vgsec

/dev/dsk/c2t6d1
/dev/dsk/c2t6d2
vgchange -a y vgsec
vgcfgbackup vgsec

Disk Array

Problem after a reverse sync

/etc/lvmtab
vgid-56789
c1t3d1

vgid-56789
c1t3d2

vgprim(pvols)

vgid-56789
c2t6d1

vgid-56789
c2t6d2

vgsec(svols)
Disk Array

/dev/vgpriv (vgid:12345)
/dev/dsk/c1t3d1
/dev/dsk/c1t3d2

Solution
export volume group
vgid-56789
c1t3d1

vgid-56789
c1t3d2

reimport volume group

vgprim(pvols)
/etc/lvmtab
vgid-56789
c2t6d1

vgid-56789
c2t6d2

vgsec(svols)
Disk Array

/dev/vgpriv (vgid:56789)
/dev/dsk/c1t3d1
/dev/dsk/c1t3d2

VxVM Concepts

VxVM Intro
Another Volume Manager Supported on HPUX platform
Can co-exist with LVM
Allows flexible disk space management
includes vmsa gui as well as cli
Available from HP beginning with HP-UX 11i.
Available from Veritas for other releases
No rootability until 11i Release 1.5 (11.20)
Lite & Full versions

VxVM Objects - Disks

Physical Disk:
Disk initialized for VxVM
use

Free disk pool

or

private region

Disk Group - dgname

cntndn or
enclosure based
name

dgname

dgname

dgname

dgname01

dgname02

dgname03

VMdisks

VxVM Concepts-Disk Group

Similar to an LVM Volume Group
One or more VMdisks
Pool of space that can then be "partitioned" into volumes
Entire group can be moved between systems using import/deport

dgname

dgname

dgname

dgname01

dgname02

dgname03

VxVM Concepts-Disk Group "activation"

There is no /etc/lvmtab file
hostid in private region is compared to hostid on system
autoimport flag indicates group should be imported at start up
deport removes hostid from private region but leaves disks in
tact
/etc/vx/volboot
host1

host1

host1

mydg01

mydg02

mydg03

host1

VxVM Concepts-Configuration
Database
Describes all objects contained in the disk group
Replicated on multiple disks in the disk group (4 by default)

Config Database

Config Database

Config Database

Config Database

VxVM rootdg
Required for VxVM to run
Does not necessarily contain the root file
system
11i: will not include the root file system
11i 1.5: will include the root file system
future releases will have choice
rootdg cannot be moved between systems
discourage including customer data

VxVM Concepts-Volume
Similar to an LVM logical volume
Analogous to a "partition"
Can contain a file system, a swap area or raw data
Represented by device files:
/dev/dsk/dgname/volname /dev/rdsk/dgname/volname

VxVM Concepts-plex
Data Plex: A copy of the data
Non-mirrored volumes will have one data plex, mirrored volumes
will have multiple data plexes
a plex can span disks
log plex: mirrored and raid5 volumes can have a log plex
vol01

vol01-01
(plex)

vol02
vol02-01
(plex part1)

vol02-01
(plex part2)

vol03
vol03-01
(plex)

vol03-02
(plex)

vol03-03
(log plex)

VxVM Concepts-subdisk
Contiguous space on a single vmdisk
entire subdisk belongs to a single plex
usually created when volume is created

volume
vol01

plex(es)
vol01-01

subdisk(s)
mydg01-01

vol02

vol02-01
vol02-02

mydg02-01
mydg03-01

vol03-01

mydg02-02
mydg03-02

vol03

mydg01

mydg02

mydg03

mydg02-01

mydg03-01
mydg03-02

mydg01-01
mydg02-02

VxVM Objects - Review

Data
vol01 (volume)
vol01-01 (plex)
dgname01-01 sub-disk

vol01-02 (plex)

dgname02-01 sub-disk

sub-disk

dgname01 (vmdisk)

dgname02

dgname03

cntndn (physical disk)

cntndn

cntndn

dgname03-01

Data Plex Layouts

Concatenated
Striped
raid5

Concatenated Plex Examples

Simple volume

Spanned Volume

Mirrored Volume

One plex

One plex

Two plexes

One subdisk

Two subdisks

Two subdisks

Striped Plex Layout

d1 d4 d7 d10

d2 d5 d8 d11

d3 d6 d9 d12

raid5 plex Layout

d1 d4 d7 p4

d2 d5 p3 d10

d3 p2 d8 d11

p1 d6 d9 d12

Mixed volumes
You can combine plex layouts in a single volume
By default the striped plex will be preferred over the concatenated
pplex
You can not mirror raid5 plex layout
mixed volume

d1 d4 d7 d10

d2 d5 d8 d11

d3 d6 d9 d12

plex1

plex1

plex1

plex2

Layered Volumes-Stripe-Mirror
1

Raid 1+0

Mirror-Stripe
1

Raid 0+1 Stripe-Mirror (Layered Volume)

1

1
Mirrored subvolume 1

2
Mirrored subvolume 2

Striped-Volume

Mirrored subvolume 3

Layered Volumes-Concat-pro

Normal Mirror (spanning multiple disks):

8.9 G

1.6G

8.9 G

1.6G

Layered Volume (spanning multiple disks):

1.6G
8.9 G
1.6G
8.9 G
sub-volume
(layout=mirror)

sub-volume
(layout=mirror)

volume with layout=concat

plex

Creating Volumes
vxassist (cli) or vmsa (gui)
you create volume
plex, sub-disks automatically created

bottom up
manually create sub-disk(s)
specify sub-disk(s) to form plex
specify plex(es) to form volume

Dynamic Multi-Pathing
/dev/vx/dsk/userdg/myvol

/dev/vx/rdsk/userdg/myvol

/dev/vx/dmp/c2t3d0

/dev/dsk/c2t3d0 /dev/rdsk/c2t3d0

/dev/vx/rdmp//c2t3d0

dev/dsk/c7t3d0

Controller(SBA) c2xxxx

/dev/rdsk/c7t3d0

Controller(SBA) c7xxxx

Multiple paths

VxVM-LVM Comparisons
LVM
Physical Volume
Physical Volume in a VG
Volume Group
Logical Volume
group of contig. extents
single mirror copy
MWC/MCR
pvra/vgra
extents

VxVM
Physical Disk
VMdisk
Disk Group
Volume
subdisk
plex
dirty region log
private region
public region

Information Sources

H6285S Hands-On with LVM course

H6487S Hands-On with MC/ServiceGuard course

H5875S HP-UX System and Network Admin for Experienced

UNIX System Administrators

U1592AAE Intro to VxVM - HP Education Virtual Classroom

Seminar
www.hp.com/education

Various white papers and cookbooks from the Response Center