Thread: Re: BUG #3245: PANIC: failed to re-find shared lock ob ject

> Von: Tom Lane [mailto:...]
> Gesendet: Freitag, 20. April 2007 18:14
> An: Dorochevsky,Michel
> Cc: pgsql-bugs@postgresql.org
> Betreff: Re: AW: [BUGS] BUG #3245: PANIC: failed to re-find shared lock
object
>
> "Dorochevsky,Michel" <michel (dot) dorochevsky (at) softcon (dot) de>
writes:
> > Two runs of the same test program fail at different places, so it seems
to
> > be dependent of the timing. Two log files are available at
> >    www.dorochevsky.de/infos/postgresql-2007-04-20_145638.zip
> >    www.dorochevsky.de/infos/postgresql-2007-04-20_144006.zip
>
> Hm, could you try again with log_line_prefix set to '%m %p ' rather than
> just %t?  It's too hard to guess which messages are coming from which
> backend ...
>
>            regards, tom lane

Took me a number of tries, here it is
    www.dorochevsky.de/infos/pg_log-2007-04-21_1811.zip

Best Regards
-- Michel

Question: do you have any leftover files in $PGDATA/pg_twophase ?

I'm wondering why the log contains no warning messages about stale
two-phase state files.  It looks to me like the system should have
found the two-phase file still there upon restart, but the transaction
should have been marked already committed.

BTW, can you tell whether the failing transactions actually were committed
--- are their effects still visible in the database?

            regards, tom lane