On Thu, 21 Jun 2007, Tom Lane wrote:
> "Henka" <henka@cityweb.co.za> writes:
> > I happened to notice this error in the log when my application was refused
> > a db connection (quite unexpectedly):
>
> > PANIC: corrupted item pointer: offset = 3308, size = 28
> > LOG: autovacuum process (PID 18165) was terminated by signal 6
>
> FWIW, the only occurrences of that specific message text are in
> PageIndexTupleDelete and PageIndexMultiDelete, so you can be pretty sure
> that this is just a corrupted-index problem. Once you've identified
> which table has the problem, a REINDEX should fix it.
I've identified the offending index, but REINDEX also causes the PANIC.
I also tried DROPping the index (and REINDEX DATABASE foo), but that too
causes a panic.
Is it possible to identify the physical file/s used by this index and
remove it manually from the filesystem (while pg is offline) and update
system tables, or is there some other way I can reindex/drop/remove it?
The only thing I haven't tried is dropping the entire table, but I suspect
that will also cause a panic.
If I can't trash the index externally, is possible to trash the entire
table/index group externally?
Any comments are appreciated.
Regards
Henry