Night shift. Same transaction started throwing -911 around 01:30 during month-end batch. SQLERRM mentions deadlock or timeout. Victim is always the online update.
Batch team says their job "only reads." I do not buy that. Where do you look first to see who held the lock?
Check DB2 lock timeout vs deadlock in the detail. -911 can be either. DISPLAY DATABASE LOCKS or your monitor (OMEGAMON / Sysview / whatever you have) around the failure time.
Also "read only" batch with UR or bad isolation can still block if it scans hot pages. Get the batch plan name from the victim's IFCID if you capture them.
We had a "report" that did SELECT FOR UPDATE by accident after a copy-paste. Worth grepping the batch source.
It was timeout, not deadlock. Batch held a tablespace lock longer than online could wait. They are splitting the job. Survived the night.
You must be signed in to reply to this thread