In some cases, we might observe messages similar to the
following indicating that the cf encountered a hang. Due to this
db2 monitoring code would receive a timeout and we will see the
cf being restarted.
2016-12-29-10.10.26.664040-120 I1646A574 LEVEL: Error
PID : 17825896 TID : 1 PROC :
db2rocme 128 [db2inst1]
INSTANCE: db2inst1 NODE : 128
HOSTNAME: test_host
EDUID : 1 EDUNAME: db2rocme 128 [db2inst1]
FUNCTION: DB2 UDB, RAS/PD component, pdLogCaPrintf, probe:876
DATA #1 : <preformatted>
Ps Get Server Role failed. status: 8006005a
DATA #1 : <preformatted>
If a CF return code is displayed above and you wish to get
more information then please run the following command:
db2diag -cfrc <CF_errcode>
....
2016-12-29-10.10.26.669247-120 I4714A522 LEVEL: Error
PID : 17825896 TID : 1 PROC :
db2rocme 128 [db2inst1]
INSTANCE: db2inst1 NODE : 128
HOSTNAME: test_host
EDUID : 1 EDUNAME: db2rocme 128 [db2inst1]
FUNCTION: DB2 UDB, high avail services, rocmCAMonitor, probe:761
MESSAGE : ECF=0x94C6005A=-1798963110
DATA #1 : String, 70 bytes
CF has encountered a hang, registering a timeout and returning
offline
DATA #2 : ZRC, PD_TYPE_ZRC, 4 bytes
0x00000000
.....
Checking the cf stacks, we will observe the CF thread holding a
lock while traversing a list of elements to free and all other
threads would be blocked waiting on this lock. The problem is
caused due to a timing window that exists when modifying this
list, which results in the list being corrupted. Due to this we
end up with a circular list that the CF thread traverses, which
thereby causes an infinite hang.
This APAR addresses the above observed issue. |
