|
Problem IC85237
|
Status: Closed |
PERFORMANCE DEGRADATION CAUSED BY LATCH CONTENTION ON SQLO_LT_SQLP_DBCB__HEADLSNSEM. |
| product: |
DB2 FOR LUW / DB2FORLUW / 950 - DB2 |
| Problem description: |
In high workload circumstance, i.e. many concurrent transaction
commit, there might be many EDUs waiting for latch
SQLO_LT_SQLP_DBCB__headLsnSem in following stacks.
sqloXlatchConflict
sqloXlatchConflict
sqlpUpdateMinBuffLsn
sqlpUpdateLowTran
sqlpScanFirstLsoArrayForLowTran
sqlpEndUowRuntime
sqlpxcm1
sqlrrcom_dps
sqlrrcom
sqlrr_commit
sqljs_ddm_rdbcmm
sqljsParseRdbAccessed
The stack trace might vary. But, due to the heavy latch
contention on headLsnSem, the performance will be impacted
significantly.
The fix for this APAR will reduce the latch contention on
SQLO_LT_SQLP_DBCB__headLsnSem. |
| Problem Summary: |
****************************************************************
* USERS AFFECTED: *
* ALL *
****************************************************************
* PROBLEM DESCRIPTION: *
* In high workload circumstance, i.e. many concurrent *
* transaction *
* commit, there might be many EDUs waiting for latch *
* SQLO_LT_SQLP_DBCB__headLsnSem in following stacks. *
* *
* sqloXlatchConflict *
* sqloXlatchConflict *
* sqlpUpdateMinBuffLsn *
* sqlpUpdateLowTran *
* sqlpScanFirstLsoArrayForLowTran *
* sqlpEndUowRuntime *
* sqlpxcm1 *
* sqlrrcom_dps *
* sqlrrcom *
* sqlrr_commit *
* sqljs_ddm_rdbcmm *
* sqljsParseRdbAccessed *
* *
* The stack trace might vary. But, due to the heavy latch *
* contention on headLsnSem, the performance will be impacted *
* significantly. *
* *
* The fix for this APAR will reduce the latch contention on *
* SQLO_LT_SQLP_DBCB__headLsnSem. *
****************************************************************
* RECOMMENDATION: *
* To reduce the chance of hitting this problem, following *
* methods *
* can help increase the frequency of flushing dirty pages. *
* - set db cfg parameter SOFTMAX to a small value less than *
* 100, *
* eg. SOFTMAX=25. *
* - specify separated buffer pool for different temp *
* tablespaces. *
* *
* OR *
* *
* Upgrade to v9.5 FP10 *
**************************************************************** |
| Local Fix: |
To reduce the chance of hitting this problem, following methods
can help increase the frequency of flushing dirty pages.
- set db cfg parameter SOFTMAX to a small value less than 100,
eg. SOFTMAX=25.
- specify separated buffer pool for different temp tablespaces. |
| Solution |
The fix of the problem is included in v9.5 FP10 |
| Workaround |
not known / see Local fix |
| Timestamps |
Date - problem reported :
Date - problem closed :
Date - last modified :
| 11.07.2012
27.08.2012
27.08.2012 |
| Problem solved at the following versions (IBM BugInfos) |
9.5.FP10
|
| Problem solved according to the fixlist(s) of the following version(s) |
| 9.5.0.10
|
 |
