>>> On Sat, Mar 25, 2006 at 8:40 pm, in message <22874.1143340808@xxxxxxxxxxxxx>, Tom Lane <tgl@xxxxxxxxxxxxx> wrote: > Jerry Sievers <jerry@xxxxxxxxxxxxxxxx> writes: >> At any rate; I'm wondering what possible causes might be responsible >> for pg_stat_activity's underlying functions to lose track of the valid >> process list? > > It sounds like the stats collector missed a few "backend quit" > messages. This isn't real surprising: the stats messaging mechanism is > intentionally designed to drop messages under severe load, rather than > slow down backends. Is there any way to tweak this in favor of more accurate information, even if has a performance cost? We're finding that during normal operations we're not seeing most connections added to the pg_stat_activity table. We would like to be able to count on accurate information there. We've been considering adding functions to get at the underlying structures to be able to retrieve it, but it would make a lot of sense (for us, anyway) to make this table accurate instead. What would be involved in that? Would it improve the accuracy of the other statistics, as well? Would anyone else be interested in something like this (probably controlled by a configuration option), or are we unique in this regard? -Kevin
