Jan Wieck <JanWieck@Yahoo.com> writes:
> Tom Lane wrote:
>> The attached patch fixes the problem by causing the stats collector to
>> detach from shared memory, which it isn't using anyway.
> I seem to recall there once was a pipe from the postmaster to the stat's
> processes and closing that will actually get rid of them.
Yeah, plan B would be to force the stats processes to die and be reborn
along with the backends. However that seems like a much larger
behavioral change than just detaching them from the shared memory.
If we ever need the stats collector to be able to access shared memory,
we'd need to do it that way; for now I'll take the simpler solution.
> Anyhow, good catch. The judgement to apply to both looks right to me.
Will do.
regards, tom lane