>>> On Thu, Apr 6, 2006 at 1:26 pm, in message
<26154.1144348014@sss.pgh.pa.us>,
> "Kevin Grittner" <Kevin.Grittner@wicourts.gov> writes:
>> Tom Lane <tgl@sss.pgh.pa.us> wrote:
>>> You weren't by any chance running with full_page_writes = off
>>> were you?
>
>> Yes we were. Apparently I have misunderstood the implications of
this.
>
> So had we all :- (. It just plain doesn't work in 8.1.*, and will
be
> disabled in 8.1.4 --- see discussion last week.
Dang! I've not been able to keep up with the lists, and I missed that
thread. I'll change that setting as soon as possible.
> Anyway, that explains your "heap_clean_redo: no block" failure. I
think
> you're stuck risking a pg_resetxlog to try to get back into the
> database. If that results in a hopelessly corrupt database, we can
try
> modifying the WAL replay code to not consider this a fatal error,
and
> see if that produces anything we can use for debugging.
Will do. Before I do that, though, is it worth making a copy? (The
down side is primarily the delay of copying it.)
> I'm glad this isn't your only copy of the database ...
We go beyond being a belt-and-suspenders shop. Think staples and
glue-gun, too. ;-)
This is one of four copies of what is redundant data to start with, and
we have backups. So recovery is no sweat, but we got emails about
timeouts from the public during this incident, so the underlying bug is
important to us.
> As long as you're comfortable doing stuff like pg_filedump and
modifying
> the code to get more debug info, we can proceed without getting into
the
> question of remote access.
OK. If you later think we do need to go that direction, send me an
email off-list.
-Kevin