1812670 - CO_ITEM: Long runtimes. ‘Component CO-OM (Overhead Gost Contig), Version 4 Released On 15.05.2017 1 symptow You archive co Tine itens that have the archiving abject co_37EH and you notice inexplicably Tong runtines for the archiving programes that are involved. I other Terms 20, Archive Development Kit, COLITIM, COLITEW MRI, Co_ITEALOEL, RARCEOAL, RARCCOA2, COUP | Reason and Prerequisites here are several | solution Reasons for Tong runtimes could be the following $n parctcular + oatabase problens + the volune of data + the custonizing and/or the selection eritersa If you suspect database problens (for example, because other prograns that access the Tine ‘tens run for a disproportionately Tong avount of time), an SQL trace nay provide “nforeation about whether your suspicion 4s correct. 1 this is the case, adeitioral analyses are required on the database, For sata volumes, che number of entries in the table COEP, as nell as the nunber of co objects that are involved (COEP-ORIMR; this, could be the following: Cost centers, orders, projects, business processes, networks, and so on (in short: all objects chat can be assigned in co)). play an important role. If data volumes are very high, many CO objects are involved. and a very large sataset 1s to be archived, you mast also sesime that runtine will be correspondingly Tong and we cannet provide an estinate of how Tong this wiTl cake here However, if runtime stiT7 appears to be too high, fursher analyses are required. To do this, you mist examine the data volume, the Customizing, and the selection criteria in more detail before you contact SAP You can use the programs RARCCOAI (ereate statistics) and RARCCOA? (display statistics) to determine the data volue. Ta the header of AARCCON2, you See when the statistics were created with RARCCOAL. They should be as current as possible but should at least ‘include the archiving period, For nore information about the progrars, see sa note 134688 Note that RARKCOAZ SE1TT recognizes che obsolete objecc CO_COSTCTR (See SAF Note 868671). However. in connection with CO_TTEM, it 4s only relevant that all object types £o which an archiving abject is assigned in the statistics can also be archived with COLITOA. You can use the program AKOBSHON to determine how many CO objects of each object type exist for a controlling area. You must almays Stare REOWSHON and RARCCOAL in the Background Because creating the statistics for certain abject types can be very tine consuming Co100H archives the intersection of Tine Stess that 4) are alToned for archiving in accordance with customizing (transaction AKCULCOITL, refer to SAP nove 73152 and the FL help for the indivieual settings) anct b) that are to be archived in accordance with the accrual calculation on the selection screen of the write progr. SAP Notes 79152 and 1414012 deseribe how this is done in detail among other things. Xt is iaportant that the systen first detersines the co objects for each object type involved and then the relevant data for each object, therefore, the main selection criterion 4s the object and additional accrual calculations (to-fiscal year, to-peried, value types) are subaccrual calculations that have Tittle fefore you start CoLITEM archiving, you mist therefore think aboue how the contents are structured in your archive: in particular, with regard to later archive evaluations, For nore information, see SAP Note 1414012 rom the point of view of runtize aptimization, you must consider the following above al1: 1 an the neite program, ‘include only those co objects that are actually to be archived and for which Tine iene exist in ‘the archiving period. To do this, the object type. subobject type (only for orders). group, or set selection criteria, and the application Custonzing (transaction ARCULCOTTD) are provided,‘TF you Teave the object type blank in the write program, the system processes all object types that are allowed in Customizing. You can use RARCCOAZ to see the object types for which Tine tens were actuaTTy posted. Fron this, you may detereine that wary objects (See AKOBSHO without data exist for certain object types; these would then put unnecessary scrain on archiving Therefore, in customizing, you should aTTow only those object types that actually have Tine itens in the archiving periog and/or use the object type as a selection eviterion in the write pragran to exclude blank object types fron archiving 2. The oroup oF Set" criterion is well docurented (F1). Grouping almays makes sense if many objects exist in an object, ‘ype but most of these are enpty (thar 45, they do nor contain any Tine itexs) xample: only 1,000 of 5,000,000 sales document objeces contain Tine ‘tens. Performance inproves if you create a group ‘chat containg these 1,000 sales document objects and archive this rather than archiving al? §,000.000 sales docurent objects that exist Grouping is also suitable for object types that have Targe datasets to portion if the renaining selection criteria are not sufficient for this ‘The Following section answers sone Frequently asked questions about performance analyses for COTTE: question: How can 1 Find out the criteria with which coLrTEK war was started? Ansner call transaction SARA for Co_ITEM, navigate to adinistration, position the cursor oh the archiving run that interests you fand righteclick to cal? the user entries If you are using the job overview in transaction M37, place the cursor on the write job that interests you and choose che option co ‘isplay the step Hist, Tn the step Hist overvien, choose "Goto => Variant” to navigate to the display of the relevant variant. question: How can 1 start a perfomance trace for awrite job that 4s running? ‘ansner C211 transaction STOS and select the “Activate Trace with FiTzer” options Were, you have various options to restrict the trace to the write job that is runting, Allow the trace to run for a while: then deactivate ft again anc ctaplay the results, Database accesses with poor performance are marked in red, In addition, the SELECT statenent (For exanple) contains the object nueber hae was Just used (Field OB)NR). The first two characters specify the language-indesendent internal abject type. Decoding Information 15 provided in the table 12002, question: wy write job uses several objects in the sane run. How can x deterwine the object type that the job is currently processing? Ansmers Start a performance trace that must comprise of at Teast one SELECT on the Tine ftens table, Tn ORIARAD(2), you can S60 the Internal object tye Alcernatively. you can use the job Tog and the resules of RKOESHOW to estimate this. The (internal. Tanguage- independent) object types are processed alphabetically, The Job Tog displays the nusber of objects processed up to nom every 30 minutes, Together with the results of Reo#show that are cumslated line-by-Tine (25 long as the write Job is restricted to one controlTing area), you e2n use this to determine che object type that was processed a the tine Of the TASt og entry, (question: The Jeb has already been running for a long tine: should 1 cancel ie? ‘nsmer Use RKOWSHOW to detentine the variant of the wrive job and the custosizing of how many objects in the job thar is runing are to be processed. use the job Tog to check how any objects have aTready been processed. in this way. you can estinate the Fenaining runtine, I the system has recently reached the end of the job, you must not cancel the job (untess other reasons require che) Frage: the jab 4s running for a disproportionately long tine. Hom can x obtain additional information about what the job 1s currently doing? Ansner Navigate to the process overview of the server on which the write job is running (Eransaction sw8i/SW50). Then choose F& (afresh) to see how quickly the display changes (table accesses, internal program processing procedures), This enables you to draw conclusions on the progran behavior ané performance ang can be the starting point for further analyses. For example, iF the Job remains in certain aragran parts for a disproportionately long tine without the system accessing the database, this may indicate source code that 1s not optinal ‘question: why 6oes the job run for so Tong even though no Tine itens or very few Tine stens exist according to AARCCON? ‘nsmers Runtine mainly depends on the runber of objects that must be processed in a run, ror each obfect, the system uses the Selection criteria to check nhether data exists. only after st daes this do you receive the result. 1f many objects are involved even if they are empty), thelr processing Still requires the corresponding anount of elm ‘question: 1m conection with performance investigations, 1 require statistics for tables that do not exist in RARCCOA2, How can I create these? dnsmer: table analysis (transaction TAMNA) provides numerous options for statistical table evaluations based on field value Combinations. As a result, you can also use virtual flelds to evaluate parts of flelds. this can be very helpful for different analyses. Brample: You can use an ad-hoc variant for the table CoEP with the Fields KOKRS, GIAMR, and (virtually) OBINR+D(2) to create siwtlar‘statistics to those im RARCCOAZ. Alternatively, an ad hoc variant for the table AUFK with the fields AUART and AUTYP would deliver Information about how the total number of orders is sp1it between the combination oF orger type and order category ‘question: archiving runs for a very Tong tine, the end is not in Sight, and 1 suspect errors. How can 1 determine whether archiving with the current selection erfteria and settings contains errors? ‘ansmers create a group that contains a manageable nunber of co objects that have the affected object type. cal? transaction E28 for Co_TToLHtE in 2 test run, on the Selection screen, use the group That you Just created ard the accrual calculations of the archiving process that 4s not coming to anend, The runtiee of the test and its result provide information about whether runtine probTens actually exist or wether the problen Ties elseshere, This type of test run can also be useful to use éeduaging to determine the ‘cause of runtine problens. ove to existing locks, this test run may not be possible until after an archiving process that is cunning ‘im paralTel 1s completed or canceled. ‘question: The write Job terminates with a memory overFlon. can I prevent this? rnsmer: A common cause 1s the settings of the write progras of Co_LIMM in relation to Togginge For each CO object that was processed, the syster writes a Tog entry: this 45 either a success wessage or an error message (no data) ‘he number of objects that were processed in retained in the log in intervals of 30 minutes. 7f the number ix very high at the tine of the tersination, detailed logging is probably active. You can use the relevant variant to fSnd this out. (See above.) If detavled Tagging ‘5 active, read the 1 documentation for the “oetatl Log” carefully on the selection screen af the write program. re articular, note the following section! “only generate detail logs for snall asounts of eata, for example \f you are working in the ‘est mode dnd you are processing few objects. Otherwise the prograr may terainate, because the Toad on the menory 18 too large.” | Manual Activities | This document is referenced by