Re: pg_clog

Thanks, I tought about that as well but with PG7.X we had the same machine
with the same database running smoothly for more than a year
now strange things like this are happening:

2006-12-18 14:49:35 CET [14691] : LOG: server process (PID 13421)
wasterminated by signal 11
2006-12-18 14:49:35 CET [14691] : LOG: terminating any other active
serverprocesses
2006-12-18 14:49:35 CET [14009] db idle: WARNING: terminating connection
because of crash of another server process
2006-12-18 14:49:35 CET [14009] db idle: DETAIL: The postmaster has
commanded this server process to roll back the current transaction and exit,
because another server process exited abnormally and possibly corrupted
shared memory.
2006-12-18 14:49:35 CET [14009] db idle: HINT: In a moment you should be
able to reconnect to the database and repeat your command.

I bet a crashing process could cause that data corruption, right?

On 12/18/06, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
>
> "Edoardo Ceccarelli" <eddy(at)axa(dot)it> writes:
> > pg_dump: ERROR: could not access status of transaction 1629514106
> > DETAIL: could not open file "pg_clog/0612": No such file or directory
>
> This looks like a data corruption problem to me. It may or may not be
> significant that 1629514106 = 0x6120697a = ASCII 'zi a' (or 'a iz' if
> you are on a big-endian machine). That looks at least a little bit
> plausible for Italian text, so I'm wondering if your disk drive or OS
> has hiccuped and dumped a sector or two of some other file into your
> database :-(. We've seen it happen before.
>
> Check the archives for past discussions about locating and recovering
> from corrupted data.
>
> regards, tom lane
>