Klaus Ita wrote:
>>> I have restored a Database Cluster with a recovery_target_time set to
>>>
>>> recovery_target_time = '2013-07-27 21:20:17.127664+00'
>>> recovery_target_inclusive = false
>>>
>>>
>>>
>>> now it seems the restore rather restored to some point in time (rather the 18th than the 27th). Is
>>> there an explanation for this huge gap? Is that the last 'consistent state'?
>>
>>
>> Maybe the log entries created during restore can answer the question.
> 2013-07-30 11:15:15 UTC <%> LOG: starting point-in-time recovery to 2013-07-27 21:20:17.127664+00
> 2013-07-30 11:15:15 UTC <%> LOG: restored log file "00000001000002300000005C" from archive
> 2013-07-30 11:15:15 UTC <%> LOG: restored log file "00000001000002300000005A" from archive
> 2013-07-30 11:15:15 UTC <%> LOG: redo starts at 230/5ACD7CC0
> ...
> ...
> ...
> 2013-07-30 14:28:45 UTC <%> LOG: restored log file "000000010000026400000002" from archive
> 2013-07-30 14:28:45 UTC <%> LOG: unexpected pageaddr 263/C706C000 in log file 612, segment 2, offset
> 442368
> 2013-07-30 14:28:45 UTC <%> LOG: redo done at 264/20698A8
> 2013-07-30 14:28:45 UTC <%> LOG: last completed transaction was at log time 2013-07-18
> 11:42:22.121512+00
> 2013-07-30 14:28:45 UTC <%> LOG: restored log file "000000010000026400000002" from archive
> cp: cannot stat `/var/tmp/xlogs_recovered_2013-07-30/wal_files/00000002.history*': No such file or
> directory
> mv: cannot stat `/tmp/00000002.history': No such file or directory
> 2013-07-30 14:28:45 UTC <%> LOG: selected new timeline ID: 2
> cp: cannot stat `/var/tmp/xlogs_recovered_2013-07-30/wal_files/00000001.history*': No such file or
> directory
> mv: cannot stat `/tmp/00000001.history': No such file or directory
> 2013-07-30 14:28:45 UTC <%> LOG: archive recovery complete
> 2013-07-30 14:29:09 UTC <%> LOG: autovacuum launcher started
> 2013-07-30 14:29:09 UTC <%> LOG: database system is ready to accept connections
>
> well, that does not indicate anything for me.
To me it indicates that log file 000000010000026400000002 might be corrupt.
PostgreSQL stops replaying WAL after it detects a corrupt WAL record.
Do you have a second copy of the WAL file?
Yours,
Laurenz Albe
