Re: strange behavior (corruption?) of large production

Tom Lane wrote:
> Joe Conway <mail(at)joeconway(dot)com> writes:
>
>>So they agree in template1 and cyspec databases.
>
> OK, in that case I'd wonder about whether you've suffered XID wraparound
> in pg_database and/or pg_shadow. The typical symptom of this is that
> entries are valid from the system's point of view but not visible to
> queries, and that seems to be what you have. If so, a restart will NOT
> fix it. You could try a VACUUM FREEZE on pg_database though. Before
> doing that, I'd suggest looking at the rows' xmin values (use
> pg_filedump or grovel through the bits by hand) to confirm the
> wraparound theory.
>

Talking to the maintainer of this cluster, it sounds like XID wraparound
could be the problem. I thought they were running database wide vacuums
at some regularity, but apparently they are only vacuuming specific
production tables.

Since this is a production machine, putting pg_filedump on it may be
problematic -- if I grovel through the bits by hand, can you give me a
hint about what to look for?

Thanks,

Joe