Louis-David Mitterrand <vindex@apartia.org> writes:
> [7.3.2 on debian unstable]
> We recently had an unclean server shutdown and now a pg_dump command
> crashes the backend:
> pg_dump: PANIC: open of /var/lib/postgres/data/pg_clog/0003 failed: No
> such file or directory
> Should we restore from a known good dump after an initdb?
If you have a good dump then initdb and reload seems like the best
answer. Before you do, though, I'd be interested to see an ls -l of the
pg_clog directory (I'm interested in the names/sizes/mod times of the
files therein). Can you give any more details about the server
shutdown?
regards, tom lane