on 09-27-2012 12:17 PM
Hello All,
I am running BDLS after system copy on QA system. I am able to complete all other conversions for A* to Z* tables except C*.
Before running BDLS for C* I analysed previous BDLS logs and created index on filed (LOGSYSO,LOGSYSP) with the belwo mentioned command. I have done that for other tables according to there logsys filed.
create index sapr3.coep_bdls on sapr3.coep (LOGSYSO,LOGSYSP);
Previous BDLS log shows me following entry for COEP table
COEP LOGSYSO 0 0 |
LOGSYSP 0 0 |
All other conversions went well but the conversion for C* tables is running from about 10 hours inspite of creating required indexes for all C* tables.
I can see in sm50 that it is stuck on COEP table and doing sequential read. Below is the details of teh WP running that conversion.
Database | Number | Time (usec) | Recs. |
Direct Read | 66 | 261.979 | 15 |
Sequential Read | 107 | 1875031593 | 6 |
Insert | 0 | 0 | 0 |
Update | 0 | 0 | 0 |
Delete | 0 | 0 | 0 |
Sources | 0 (Bytes) | ||
RSQL | 15.160 (Bytes) | ||
Commit | 0 | ||
DB Procedure Calls | 0 | 0 |
I am not able to understand why COEP is taking much time as other tables with almost same number of records went quickly.
Do I need to create other index as well on COEP table. Please help me.
One more question, If someone suggest me to create other index as well, so can I stop the currently running conversion and start again after applying your suggestions.
Regards
Shubham
Hi,
Refer to this doc:
http://wiki.sdn.sap.com/wiki/display/ABAPConn/BDLS+Performance+Problems
Specific conversion for large tables.
It is also possible to exclude tables from BDLS or to convert specific tables in parallel (e.g. COEP). To do this, you can either do it maually by making the required entries in the table "BDLSEXZ" as specified in note 932032
Regards,
Ashutosh
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
I have a BDLS run which so far in total will run over 150 hours for test mode. I have to do 3 runs (changing ECC Logical System, BW Logical System and also APO Logical system....so will run for about 6 weeks!!!!).
I have done above advice to running old version and breaking down A*, B* etc and have identified 9 tables which are taking up long runtime (including COEP, GLPCA, VBAP, FAGLFLEXA etc etc
For most of these tables I have been able to do individual test runs, and they run for between 2 hours and 24 hours per table (running in parallel), except for COEP which so far has been running for 165 hours and not yet finished.
I was going to look at adding indexes as directed, BUT looking at the log files created from these individual test runs, they show zero records to update for all but one table (VBAP). I was therefore thinking that I could do an update run for all tables excluding the 9 large tables (which takes about 2 hours) and then just do a separate run for VBAP, i.e. I can ignore the other 8 tables as it seems to me that they have LOGSYS but no values posted (or at least none of the values I want to change)
What do you think?
Cheers
SImon
Please create 2 indexes on COEP as follows.
create index sapsr3.COEP_BDLS_1 on sapsr3.COEP (LOGSYSO, MANDT) ;
create index sapsr3.COEP_BDLS_2 on sapsr3.COEP (LOGSYSP, MANDT) :
Also collect statistics on the created indexes.
This should be fine and speed up the conversion .
Regards
Ratnajit
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
User | Count |
---|---|
101 | |
13 | |
13 | |
11 | |
11 | |
7 | |
6 | |
5 | |
4 | |
4 |
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.