If there are many nodes, e.g. more than 90 nodes, in a DPF
environment, then it is possible for db2stop to hit the default
START_STOP_TIME timeout of 10 minutes, which would cause DB2 to
issue a kill underneath to all nodes.
Symptom:
1.) db2diag.log would have logs like the following:
2010-08-19-03.20.45.041184-300 I109662A299 LEVEL: Event
PID : 5910764 TID : 1 PROC :
db2stop2
INSTANCE: XXXXXX NODE : 000
EDUID : 1
FUNCTION: DB2 UDB, base sys utilities, DB2StopMain, probe:240
DATA #1 : String, 26 bytes
Stop phase is in progress.
2010-08-19-03.20.45.041799-300 I109962A314 LEVEL: Event
PID : 5910764 TID : 1 PROC :
db2stop2
INSTANCE: XXXXXX NODE : 000
EDUID : 1
FUNCTION: DB2 UDB, base sys utilities, DB2StopMain, probe:250
DATA #1 : String, 41 bytes
Requesting system controller termination.
and then many occurrences of the following messages:
2010-08-19-03.21.20.554971-300 I115791A390 LEVEL: Error
PID : 389258 TID : 772 PROC : db2sysc
0
INSTANCE: XXXXXX NODE : 000
EDUID : 772 EDUNAME: db2fcms 0
FUNCTION: DB2 UDB, fast comm manager,
sqkfSendConduit::ValidateConnectedLinks, probe:100
RETCODE : ZRC=0x8159006B=-2124873621=SQLKF_CONN_CLOSED "FCM
connection closed"
2010-08-19-03.21.39.244694-300 I116936A362 LEVEL: Error
PID : 389258 TID : 772 PROC : db2sysc
0
INSTANCE: XXXXXX NODE : 000
EDUID : 772 EDUNAME: db2fcms 0
FUNCTION: DB2 UDB, fast comm manager, sqkfTcpLink::closeConn,
probe:25
MESSAGE : Link info: node 14; type 4; state 5; session
0;activated 1
2010-08-19-03.21.39.244867-300 I117299A390 LEVEL: Error
PID : 389258 TID : 772 PROC : db2sysc
0
INSTANCE: XXXXXX NODE : 000
EDUID : 772 EDUNAME: db2fcms 0
FUNCTION: DB2 UDB, fast comm manager,
sqkfSendConduit::ValidateConnectedLinks, probe:100
RETCODE : ZRC=0x8159006B=-2124873621=SQLKF_CONN_CLOSED "FCM
connection closed"
2010-08-19-03.21.50.657934-300 I118444A362 LEVEL: Error
PID : 389258 TID : 772 PROC : db2sysc
0
INSTANCE: XXXXXX NODE : 000
EDUID : 772 EDUNAME: db2fcms 0
FUNCTION: DB2 UDB, fast comm manager, sqkfTcpLink::closeConn,
probe:25
MESSAGE : Link info: node 15; type 4; state 5; session
0;activated 1
......
2.) The stack trace would have the following pattern:
<StackTrace>
-------Frame------ ------Function + Offset------
0x09000000001174D4 __fd_poll + 0x98
0x09000000000A9AE0 poll + 0xC
0x09000000000A8968 res_nsend + 0xDA4
0x0900000000100D0C res_nquery + 0x130
0x0900000000100370 res_nquerydomain + 0x180
0x090000000010063C res_nsearch + 0x228
0x09000000000B3D1C res_search + 0xA8
0x0900000000106088 ho_byname2 + 0x13C
0x09000000001210E0 ho_byname2 + 0x1AC
0x09000000000A6550 gethostbyname2 + 0x190
0x09000000000A9E98 getaddrinfo2 + 0x384
0x09000000000AB2C4 getaddrinfo + 0x36C
0x0900000009515220 sqloPdbTcpIpGetAddrInfo + 0x13C
0x090000000C2DACE0 sqloPdbTcpIpResolveHostName + 0x1C4
0x090000000C2DB030 sqloPdbTcpIpResolveHostName@glue557 + 0x7C
0x09000000086A7324 sqloReadDb2nodes + 0x8C4
0x090000000836A978 RefreshDb2nodesCache__19sqkfFastCommManagerFv
+ 0x210
0x090000000834FF48
RefreshNodesInfo__15sqkfSendConduitFP14sqkfDataTargetiPb + 0x74
0x090000000928A560
CheckForFailoverConnectRetry__15sqkfSendConduitFsPi + 0x328
0x090000000927F640 HandleConnectLostEvent__15sqkfSendConduitFUl
+ 0x160
0x090000000927D6D0 RunEDU__15sqkfSendConduitFv + 0x264
0x0900000008A46E2C EDUDriver__9sqzEDUObjFv + 0xF8
0x0900000008A4C778 sqloEDUEntry + 0x278
</StackTrace> |
****************************************************************
* USERS AFFECTED: *
* Large DPF systems *
****************************************************************
* PROBLEM DESCRIPTION: *
* If there are many nodes, e.g. more than 90 nodes, in a *
* DPFenvironment, then it is possible for db2stop to hit *
* thedefault START_STOP_TIME timeout of 10 minutes, which *
* wouldcause DB2 to issue a kill underneath to all nodes. *
****************************************************************
* RECOMMENDATION: *
* Update to Version 9.5 Fix Pack 7 *
**************************************************************** |
