Florian Schaetz <Florian.Schaetz@Optitool.DE> writes:
> I've got a little problem with my postgresql/postgis server here: Sometimes, temporary tables (sessions?) seem to
"survive",which in the end leads to a shared memory error, that can only be resolved by restarting (for a moment). This
stopsthe shared memory error for a moment, but the sessions (and with them, the tables) are still there, even if there
aren'tany open connections left.
What PG version is that? Can you create a reproducible test case?
(It doesn't have to be "reproduces every time"; something like "run this
program long enough and it'll eventually fail" would be fine.)
We've heard a few previous reports of such things, but nobody ever
had it happen regularly enough to have a chance of isolating the cause.
So it sounds like you're doing something that increases the probability
greatly, and if you could bottle that, it'd be really helpful ...
regards, tom lane