Yes I suppose that was the problem...the harwdare failure was responsible
for the corrupted pg_attribute.
But I was suprise because I had a RAID 0/1 and the data got corrupted on
the second disk as well.
I am now trying to restore things from last backup and transaction logs.
Thanks all.
Regards
Bruno Cortes
> Alvaro Herrera <alvherre@commandprompt.com> writes:
> > Bruno Cortes wrote:
> >> This was even worse than the previous error. Right now I cannot enter
> >> the database neither make a dump. When I execute a command such as
"psql
> >> -U postgres center-01" I get the error:
> >> psql: FATAL: catalog is missing 10 attribute(s) for relid 1262
>
> > Sounds very much like you haven't been doing the required vacuum on all
> > system catalogs.
>
> I don't think so --- that error is coming out of the relcache/syscache
> layer, and that code uses SnapshotNow; so it's only going to care
> whether the tuples are marked committed, not whether they've wrapped
> around.
>
> I'm afraid that Bruno has lost some pages of pg_attribute (eg, they went
> to zeroes). That's not too surprising given his previous reports
> suggesting hardware problems.
>
> If the pages in question held only rows for system catalogs then it
> might be possible to copy correct values from elsewhere, but if they
> held rows for user tables then he's got serious problems :-(. My advice
> would be to start thinking about restoring from backup.
>
> regards, tom lane
>
--
