Re: Shutting down a warm standby database in 8.2beta3

On Thu, Nov 23, 2006 at 01:18:06PM -0500, Stephen Harris wrote:
> Hmm, on Linux I did a kill -QUIT on the process group and then rather
> than cleanly exiting my recovery script I "kill -9"d it. The database
> still completed recovery as previously described.

OK, I just did the same tests on the CVS tree, rather than beta3 and
I see different results. This time recovery aborts with a PANIC. I
did a "pg_ctl stop -m i" and then kill -9'd the recovery script and:

LOG: restored log file "000000010000000000000006" from archive
Looking for 000000010000000000000007
LOG: received immediate shutdown request
/tmp/rr: line 12: 22031 Quit sleep 5
PANIC: could not restore file "000000010000000000000007" from archive: return code 9

Similar behaviour was seen with "smart" and "fast" shutdowns.

Looking for 000000010000000000000005
Restoring 000000010000000000000005 to pg_xlog/RECOVERYXLOG
LOG: restored log file "000000010000000000000005" from archive
LOG: checkpoint record is at 0/5000020
LOG: redo record is at 0/5000020; undo record is at 0/0; shutdown FALSE
LOG: next transaction ID: 0/601; next OID: 24577
LOG: next MultiXactId: 1; next MultiXactOffset: 0
LOG: automatic recovery in progress
LOG: redo starts at 0/5000068
Looking for 000000010000000000000006
Restoring 000000010000000000000006 to pg_xlog/RECOVERYXLOG
LOG: restored log file "000000010000000000000006" from archive
Looking for 000000010000000000000007
LOG: received fast shutdown request
PANIC: could not restore file "000000010000000000000007" from archive: return code 9
LOG: startup process (PID 22246) was terminated by signal 6
LOG: aborting startup due to startup process failure

Restarting the database appears to go back to asking for archive log 5
again. I'll have to do some testing with more archives to see how far back
we have to go.

--

rgds
Stephen