Re: DB/clog corruption - Mailing list pgsql-general

From Tom Lane
Subject Re: DB/clog corruption
Date
Msg-id 7839.1121089262@sss.pgh.pa.us
Whole thread Raw
In response to DB/clog corruption  ("Reid Thompson" <Reid.Thompson@ateb.com>)
List pgsql-general
"Reid Thompson" <Reid.Thompson@ateb.com> writes:
> $ pg_dump -f table_dump.sql -t table_with_error dbname
> FATAL 2:  read of clog file 1559, offset 0 failed: Success

You've got a clobbered tuple header in that file (ridiculous xmin or
xmax value).  Alternatively, given the evident age of this server,
it could be a clobbered page header leading to the system following
a bogus pointer to a tuple header.

I think the track you want to pursue is identifying which page contains
the bad data and looking at it with pg_filedump to see if there's
anything recoverable there at all.  If not, just zeroing out the page
is probably the easiest way of getting to a dumpable state.  You've
probably lost this particular tuple in any case, you might or might not
have lost the whole page.

Once you get out of this, you might want to think about updating to
something newer than 7.2 ...

            regards, tom lane

pgsql-general by date:

Previous
From: wayne schlemitz
Date:
Subject: install problem
Next
From: "George Woodring"
Date:
Subject: Foreign Key written as a trigger