ADM11002E & DB2_FMP_COMM_HEAPSZ
2007-04-02-15.53.57.734000+480 E1021855H589 LEVEL: Error
PID : 5448 TID : 4648 PROC : db2syscs.exe
INSTANCE: DB2 NODE : 000 DB : SKERP
APPHDL : 0-657 APPID: AC107318.5B04.070402075313
FUNCTION: DB2 UDB, routine_infrastructure, sqlerGetFMPIPC, probe:70
MESSAGE : ADM11002E Insufficient shared memory available for communication
with the db2fmp process. Use the DB2_FMP_COMM_HEAPSZ registry
variable to increase the amount of shared memory available for fenced
routines.
2007-04-02-15.53.57.734000+480 I1022446H495 LEVEL: Severe
PID : 5448 TID : 4648 PROC : db2syscs.exe
INSTANCE: DB2 NODE : 000 DB : SKERP
APPHDL : 0-657 APPID: AC107318.5B04.070402075313
FUNCTION: DB2 UDB, routine_infrastructure, sqlerGetFMPIPC, probe:60
RETCODE : ZRC=0x8B0F003B=-1961951173=SQLO_NOMEM_UND
"No memory available in 'Undefined Heap'"
DIA8300C A memory heap error has occurred.
4 comments:
IY54160: Shared memory segment for db2fmp process is limited to 256M even db2 instance is 64 bit on AIX
Local fix
Reduce ASLHEAPSZ can reduce the communication size between
agents and db2fmp processes
Problem summary
Shared memory segment for db2fmp process is limited to 256M even
db2 instance is 64 bit on AIX
IY96625: THREADSAFE JAVA STORED PROCEDURE CAUSES MEMORY LEAK IN DB2FMP PROCESS.
This memory leak can occur only when the number of db2fmp (java)
threads exceeds the FENCED_POOL setting and is platform
independent.
Local fix
possible workaround is to increase the FENCED_POOL setting so as
to avoid the destroying of the db2fmp (java) thread when a SP
call is finished.
Yonghang: or set keepfenced=NO when performence is not a big issue.
Summary:
The threadsafe Java SP memory leak problem is fixed in FP14. So, suggest to update to FP14.
If not, suggest:
1) aslheapsz up to 256, suggest 128
2) if process management workload is too heavy, then enlarge FENCED_POOL. Or set KEEPFENCED=NO.
a tip, keepfenced has very big impact for applications in which fenced SP used.
发表评论 Add/View Comments