AIX 5.3 64 bit
DB21085I Instance "arcproda" uses "32" bits and DB2 code release "SQL08027"
with level identifier "03080106".
Informational tokens are "DB2 v188.8.131.52", "s061108", "U810098", and FixPak
Product is installed at "/usr/opt/db2_08_01".
We have started to have some issues with running our incremental backups to TSM tape, this is all controlled by Tivoli Workload Scheduler. The TWS job starts and the backup begins and the session to TSM is initialized. However the session to TSM is lost on the TSM server and the job is still running in TWS and the db2 process is still running on the server. It looks as though db2 is gathering the information to send to the TSM server but when it is ready to send the data the session is lost and the job finally abends/stops.
My question is we have seen that it takes awhile for the job to abend in tws i.e. db2 first initializes the session tsm (which then becomes lost) but the db2pb process for the backup still runs for quite some time before it then tries to send data to the TSM server which is technically no longer there and the job fails finally.
Not sure if I have been clear or not but hopefully someone will understand what I am trying to convey.
Is there a way to kill the process "nicely" so that it doesn't leave the tables/tablespaces in a funky state?
Thus far IBM has not been able to tell us anything of value via our PMR, I was hoping maybe someone else has had similar problems.
Our database is almost 2TB in 3 partitions on 2 physical servers
Increase the TIVOLI timeout to 18000(5 hours); use 3 or 4 sessions; increased buffer sizes(increase the UTIL_HEAP_SZ correspondingly); backup is writing a marker in page header that meets its criteria; has no effect on state of tablespace; you can kill the backup session any time with no adverse results.It may be reading a lot of pages before it finds a recently modified page;so for hours you may not have a "write".