When we have a HADR system, and the standby is offline. If the
database in the primary is not activated, the first connection
to the database will be the one to activate it. If we run db2
connect to <database name> or db2 start hadr on <database name>
as primary, but without the "by force" option, the connection
will try to start hadr and connect to the standby, timing out
eventually after the HADR_TIMEOUT setting and getting SQL1768N
Unable to start HADR. Reason code = "7".
2010-02-03-04.29.45.699578+000 I2838967A388 LEVEL: Warning
PID : 606706 TID : 1 PROC :
db2agent (FRH) 0
INSTANCE: db2frh NODE : 000
APPHDL : 0-171 APPID: *LOCAL.db2frh.100203042357
AUTHID : DB2FRH
FUNCTION: DB2 UDB, High Availability Disaster Recovery,
hdrEduStartup, probe:21151
MESSAGE : Info: HADR Startup has begun.
2010-02-03-04.30.16.734635+000 I2842396A552 LEVEL: Error
PID : 999878 TID : 1 PROC :
db2hadrp (FRH) 0
INSTANCE: db2frh NODE : 000
FUNCTION: DB2 UDB, High Availability Disaster Recovery, hdrEduP,
probe:20390
MESSAGE : HADR primary did not establish connection with standby
within timeout
and will shut down. BY FORCE option required to start
primary without
standby. Timeout seconds =
DATA #1 : Hexdump, 4 bytes
0x07800001D52CD008 : 0000 001E
2010-01-14-06.35.07.374445+000 I5172086A471 LEVEL: Error
PID : 1077874 TID : 1 PROC :
db2agent (AB7) 0
INSTANCE: db2ab7 NODE : 000
APPHDL : 0-8 APPID: *LOCAL.db2ab7.100114063322
AUTHID : DB2AB7
FUNCTION: DB2 UDB, High Availability Disaster Recovery,
hdrEduStartup, probe:21300
MESSAGE : HADR EDU sqlcode:
DATA #1 : Hexdump, 4 bytes
0x000000011121526C : FFFF F918
....
2010-01-14-06.35.07.374514+000 I5172558A419 LEVEL: Severe
PID : 1077874 TID : 1 PROC :
db2agent (AB7) 0
INSTANCE: db2ab7 NODE : 000
APPHDL : 0-8 APPID: *LOCAL.db2ab7.100114063322
AUTHID : DB2AB7
FUNCTION: DB2 UDB, base sys utilities, sqledint, probe:230
DATA #1 : Hexdump, 4 bytes
0x000000011121526C : FFFF F918
....
If many of the connection attempts are issued, they will all be
serialized until the database is activated:
1. The currently active connection that is trying to start HADR
is holding the database latch. The application is waiting to
reach the HADR timeout.
2. All other connections that are trying to start HADR are
queued up behind the database latch in a serialized fashion.
If in this scenario we run db2stop force, this might take a long
time, depending on how many connections have been queued to
activate the database (they will all fail with HADR timeout
SQL1768N)
When "db2stop force" kicks in, it will detect the number of
applications that need to be forced:
FUNCTION: DB2 UDB, base sys utilities,
sqeAppServices::ExecuteStopForce, probe:1000
DATA #1 : String, 47 bytes
[Force]->Number of applications to be forced :
DATA #2 : Hexdump, 4 bytes
0x0FFFFFFFFFFFD698 : 0000 0004
....
It will until all queued up applications respond, and only then
the database is actually stopped. This might take a long time,
and could be perceived as db2stop force actually being hung. |
