An operation requiring a global buffer pool purge has been
issued, for example a tablespace/table/index drop, temporary
table truncation, REORG, LOAD, etc... The purge request has been
sent by the coordinator partition to all the subordinator
partitions. The subordinator partitions have started purging
buffer pool pages.
If the buffer pool purge operation is interrupted at the
coordinator (e.g. user interrupt, node failure), the coordinator
responds immediately by rolling back the transaction and
releasing all previously acquired locks. However, the
subordinators are usually slower in responding since a buffer
pool purge may be a long running operation. During this time
frame, if another operation concerning the same object is
initiated on a subordinator again, this may result in data
inconsistency as the coordinator is currently not holding any
locks while the subordinators have not finished their
processing.
The user symptoms of this error are not known yet, the problem
has been observed during internal product testing during a DROP
TABLESPACE operation. The purpose of this APAR is to make a
global buffer pool purge uninterruptible until the point when
all subordinators have finished their current processing, and
they are ready to accept the next request. |
****************************************************************
* USERS AFFECTED: *
* All *
****************************************************************
* PROBLEM DESCRIPTION: *
* See Error Description *
****************************************************************
* RECOMMENDATION: *
* Upgrade to DB2 for Linux, Unix, and Windows 9.8 Fix Pack 5. *
**************************************************************** |
