On Wed, 2007-06-06 at 13:26 -0700, Ben wrote:
> Hey guys, I'm burning in some new hardware and just reached the
> pull-the-plug phase of testing. On restart, I see this:
>
> 2007-06-06 12:04:00 PDT LOG: database system was interrupted at 2007-06-06 11:53:56 PDT
> 2007-06-06 12:04:00 PDT LOG: checkpoint record is at 24/C29ED068
> 2007-06-06 12:04:00 PDT LOG: redo record is at 24/BEF3C3A8; undo record is at 0/0; shutdown FALSE
> 2007-06-06 12:04:00 PDT LOG: next transaction ID: 0/11773980; next OID: 24576
> 2007-06-06 12:04:00 PDT LOG: next MultiXactId: 1; next MultiXactOffset: 0
> 2007-06-06 12:04:00 PDT LOG: database system was not properly shut down; automatic recovery in progress
> 2007-06-06 12:04:00 PDT LOG: redo starts at 24/BEF3C3A8
> 2007-06-06 12:18:03 PDT LOG: unexpected pageaddr 24/9D6DC000 in log file 36, segment 224, offset 7192576
> 2007-06-06 12:18:03 PDT LOG: redo done at 24/E06DB210
> 2007-06-06 12:20:21 PDT LOG: database system is ready
>
>
> That "unexpected pageaddr" line seems bad to me, especially seeing as how
> this is on a battery-backed hardware raid with drive write caches
> supposedly disabled. Should I expect to see such lines anyway after a
> sudden power loss?
Yes.
"Unexpected pageaddr" is just one of the ways that recovery can detect
the end of the log, since there isn't ever an "end of log" marker when
we crash. (Normally it would be a checkpoint.)
--
Simon Riggs
EnterpriseDB http://www.enterprisedb.com