On Fri, Apr 16, 2010 at 12:47 PM, Alexandre Leclerc <aleclerc@ipso.ca> wrote:
> Le 2010-04-16 14:18, Kevin Grittner a écrit :
>>
>> Alexandre Leclerc<aleclerc@ipso.ca> wrote:
>>
>>
>>>
>>> At some point I got:
>>> ERROR: xlog flush request AC/FBEEF148 is not satisfied --- flushed
>>> only to AC/FB9224A8
>>> CONTEXT: writing block 0 of relation 1664/0/1214
>>> WARNING: could not writing block 0 of 1664/0/1214
>>> DETAIL: Multiple failures --- write error may be permanent.
>>>
>>
>> You're not running out of disk space where that writes, are you?
>>
>
> There is 32 GB free.
>
>>
>>
>>>
>>> It looks like the vacuum command does not want to be execute. The
>>> other VACUUM is still progressing on the main database in another
>>> postgres.exe shell.
>>>
>>
>> Wait -- are you saying you're running two postgres instances against
>> the same data directory at the same time? (I sure hope not.)
>>
>
> I did. :( Shame on me. I just realised while reading doc on postgres that it
> is not made for that but only for a single instance at the time. I hope I
> did not break anything.
You've almost certainly corrupted the data store. This is why you
should always make a complete file system backup with the postmaster
stopped before you start doing rescue work.