Re: pg_clog error after crash

On Thu, Jul 01, 2004 at 06:05:17PM -0400, Tom Lane wrote:
> Sandro Dentella <sandro(at)e-den(dot)it> writes:
> > After a crash of the (frozen) system, postgreSQL issues a message:
> > 2004-06-30 17:51:39 [3165] LOG: connection authorized: user=sandro database=sia
> > 2004-06-30 17:51:39 [3165] ERROR: could not access status of transaction 4011600129
> > DETAIL: could not open file "/var/lib/postgres/data/pg_clog/0EF1": No such file or directory
>
> It looks like you've got data corruption in one of the system catalogs
> that is accessed during backend startup. This is not good :-(.
> Do you get the same no matter which of the databases in your
> installation you try to connect to? (Not that it will matter much,
> if all your important data is in this database...)

No, really I could connect to the database that had probles! I could "COPY
table TO 'file' for *each* table of the problematic db. The problem arouse
when trying to pg_dump it or when trying to access it via libpgtcl (I use
TkSQL as you can guess from the footer ;-).

After COPYing all tables I created the file that was missing, filled with
zeros according to a mail I found of a "similar" problem. PostgreSQL
complain disappeard. As far as I can say no data seems to be lost.

I'm using PostgreSQL 7.4.2 last release of Debian Sarge.

Thanks for your time.
sandro
*:-)

--
Sandro Dentella *:-)
e-mail: sandro(at)e-den(dot)it
http://www.tksql.org TkSQL Home page - My GPL work