PG Bug reporting form <noreply@postgresql.org> writes:
> We had a situation where a query that normally ran in 10 seconds instead
> never completed. No conflicts or locks could be seen. All indexes were
> valid and in place. A full shutdown and restart of the server was done with
> no effect.
> To test we took a full backup of this system and restored to an identical
> alternate system. Having done this, the alternate system ran the query in
> 10 seconds.
> Final test was to backup and restore back into the production system -
> having done this the query returned to running in 10 seconds. Do you have
> any idea what could have caused this?
The most obvious theory is a change of query plan, perhaps due to having
up-to-date ANALYZE statistics in one case and not the other. I don't
suppose you captured EXPLAIN output for the non-working state?
regards, tom lane