Alan McKay <alan(dot)mckay(at)gmail(dot)com> writes:
>> pg_locks? Somebody taking exclusive lock on a widely-used table might
>> explain that.
> OK, in theory we could do the following, no?
> Use our PITR logs to restore a tertiary system to the point when we
> were having the problem (we have a pretty wide 2 or 3 hour window to
> hit), then query the pg_locks table on that system?
No, that wouldn't tell you a single thing. pg_locks is a dynamic view
of current system state. You would have to look at it while having
the problem.
regards, tom lane