Hi
When we are doing high volume loading, the process bottlenecks at the System Log/System Load.
This happens from what I understand because the loading to App Group (system log/load) is sequential.
So the threading I do in my shell script to ensure that at any gives point in time each thread is loading data into a different app group doesnot help to improve loading performance as it would anyway bottleneck at system log/load app group.
Then I suppressed all the messages that are logged to system log.
But I still see 87, 88 and 329 message numbers.
Is it possible to completely stop this logging?
Note that we are doing historical data loading and would switch on loading when we start day forward stuff.
Thanks
Pankaj.
See it tool 12 seconds to load this small file.
05/31/2013 03:06:13 ADMIN 37840 Info Yes 87 Application Group Load: Name(Loans) LoadId(5162-0-0-17FAA-15840-15840) File(/cfsc/RM/scripts/tempDir/21823576/pcit0030.alpha.20130514.20130514.121212.927679.batch.rpt) InputSize(10760) OutputSize(2016) Rows(1) Time(11.8640) Appl(xxxxxxxxxxxxxxxxx)
05/31/2013 03:06:13 ADMIN 37840 Info No 32 Logoff
05/31/2013 03:06:07 ADMIN 37840 Info No 82 Application Group Object Store: Name(Loans) Agid(5162) NodeName(-CACHE-) Nid(0) Server(-LOCAL-) ObjName(17FAAA) Time(0.003)
05/31/2013 03:06:07 ADMIN 37840 Info No 86 Application Group Load Database: Name(Loans) Agid(5162) Rows(1)
05/31/2013 03:06:07 ADMIN 37840 Info No 82 Application Group Object Store: Name(Loans) Agid(5162) NodeName(-CACHE-) Nid(0) Server(-LOCAL-) ObjName(17FAA1) Time(0.002)
05/31/2013 03:06:01 ADMIN 37840 Info No 30 Login: xxxxxxxxxxxxxxxxxxxxxxxxxxxxx non-SSL (AIX) (ARSLOAD) (8.5.0.5)
I guess using arsadmin load skips the 87/88 messages in system log and 82/86 messages can be turned off from AG settings.
Are you using DB2 HADR?
You can also reduce the amount of logging per-Application Group by unchecking the logging for Object Store and Object Retrieve.
Can you paste in the output from arsload? That will tell us where the bottleneck is (indexing vs. loading).
Also, try to get a database snapshot (or read the diagnostic log) for those 12 seconds, to see if there are any deadlocks, either in the AG or the System Log.
-JD.
Hello Pankaj,
just a small question, do you use also the "arslog"?, if yes it could also have an influence in the time for logging. And maybe to "deactivate" it could also help.
Sincerely yours,
Alessandro