Saurabh Dave wrote:
>>>If your problem is something that develops over time, one thing
>>>you could watch out for are resource leaks, most prominently
>>>memory leaks.
>
> What is the recommended way to do that? It seems postgres
> uses shared memory, and we have 15 connections all taking up
> around 35MB as per task manager OR top. We are not able to
> figure out definitatively if there is a memory leak.
> Help/Clues??? Thanks all for your responses so far.
I'm just ranting, a memory leak frequently does not impact
performance (before the machine starts thrashing...).
When watching for memory leaks, you can (as a crude measure)
daily list the memory that each process consumes and
track notable increases for long running processes.
PostgreSQL shared memory should be excluded here: it does
not increase and it will show up in each backend, thus
confusing the issue.
Once things get slow, find out if the CPU or the disks
are busy and watch out for unusual "vmstat" output.
Yours,
Laurenz Albe