On Tue, Sep 14, 2004 at 08:13:24AM -0400, Matthew T. O'Connor wrote: > On Tue, 2004-09-14 at 00:46, Chester Kustarz wrote: > > On Mon, 13 Sep 2004, Matthew T. O'Connor wrote: > > > I backed up the entire data directory, and did a pg_resetxlog, but that > > > didn't help. I found the specific row that seems to be the problem, but > > > I can't delete it. > > > > I have used TRUNCATE on the table in this situation to recover. Another > > option might be to DROP the table. Or perhaps restore from backups. > > I would really prefer not to do that as pg_dump has apparently been > failing for a while so I would lose a fair amount of data.
You can create a pg_clog file (the one it's complaining about) filled with zeros, using dd if=/dev/zero bs=1k count=8 of=/path/to/data/pg_clog/0E3D and then you should be able to pg_dump the table (or at least find out if there is another corrupted tuple.) Beware that the corrupted tuple may be in there if it's supposed not to be, or it may not be if it's supposed to be. After you get your data back, I'd suggest running the usual hardware checking tools, and restore from the backup. -- Alvaro Herrera (<alvherre[a]dcc.uchile.cl>) "Use it up, wear it out, make it do, or do without" ---------------------------(end of broadcast)--------------------------- TIP 8: explain analyze is your friend
