On 31/08/2009, at 6:16 PM, Sébastien Lardière wrote:
> On 28/08/2009 18:14, Simon Riggs wrote:
>> On Fri, 2009-08-28 at 17:54 +0200, Sébastien Lardière wrote:
>>
>>> Since this moment, the slave didn't make any checkpoint.
>>>
>>> Now, we know why. Thanks a lot !
>>>
>>> But how can i fix it ?
>>>
>> Current issue: Rebuild standby from base backup.
>>
>> Cause: Locate the bug in the Index AM that causes it.
>>
>> Symptom: Currently index AMs don't detect situation that index is
>> corrupt so they keep trying to recover the index for ever, even
>> though
>> attempting to do so permanently prevents restartpoints. I would also
>> like them to have a mechanism for marking index corrupt and then
>> auto-rebuild them following recovery.
>>
>
> Thanks, I make a restore backup on the slave this morning, and It
> works !
Could you detail your solution please, if any? I've seen frozen
pg_controldata output on my standby server for ages and attributed
that to the ancient version of pgsql (8.0.x) I'm stuck with. There
have been no index related error messages in my log though.
Thank you!
Yar