The error might occur when the following criteria are met:
1) The table is an MDC table with at least two RID indexes.
2) A deferred cleanup rollout delete is performed on that MDC
table.
3) Before the asynchronous index cleanup operation completes, it
suspends due to lock timeout or deadlock, and another deferred
cleanup rollout delete is performed on the table.
4) As a result, it updates the block map incorrectly marking all
rolled-out blocks as free even though the index cleanup is not
complete. This might result in a variety of different index
errors. The problem can be detected by running the INSPECT
command with the INDEXDATA option.
You can see following output in db2diag.log when the async index
cleanup suspends(3):
2015-01-01-10.00.00.000000+540 E571009E589 LEVEL:
Warning
PID : 123456 TID : 140179515893504 PROC :
db2sysc 0
INSTANCE: db2inst1 NODE : 000 DB :
SAMPLE
APPHDL : 0-1234 APPID: *LOCAL.DB2.150101022705
AUTHID : DB2INST1 HOSTNAME: db2server
EDUID : 543210 EDUNAME: db2taskp (SAMPLE) 0
FUNCTION: DB2 UDB, AIC, aicRolloutIndexCleanup, probe:2075
MESSAGE : ADM9514I BEGIN async index cleanup on table
"DB2INST1.T1"
(ID "123") and table space "TS1" (ID "5").
...
2015-01-01-10.02.11.328114+540 I712797E573 LEVEL: Info
PID : 123456 TID : 140179515893504 PROC :
db2sysc 0
INSTANCE: db2inst1 NODE : 000 DB :
SAMPLE
APPHDL : 3-14429 APPID: *N3.DB2.150701092140
AUTHID : DB2INST1 HOSTNAME: db2server
EDUID : 543210 EDUNAME: db2taskp (SAMPLE) 0
FUNCTION: DB2 UDB, AIC, aicRolloutIndexCleanup, probe:4318
MESSAGE : ZRC=0x82A90066=-2102853530=ABP_SUSPEND_TASK_PRO
"Suspend the task processor"
CALLED : DB2 UDB, AIC, aicRolloutIndexCleanup |
