Re: pg_stat_get_backen_last_activity() ???
От | Sean Chittenden |
---|---|
Тема | Re: pg_stat_get_backen_last_activity() ??? |
Дата | |
Msg-id | 20030424000157.GX79923@perrin.int.nxad.com обсуждение исходный текст |
Ответ на | Re: pg_stat_get_backen_last_activity() ??? (Sean Chittenden <sean@chittenden.org>) |
Список | pgsql-general |
> > > Any ideas on how I can figure out what the mystery query/app/lib > > > is that's causing my backend to hang with a zillion connections > > > that are all idle? > > > > Well, that particular backend is not hung: the stack trace proves it's > > waiting for a command from its frontend. Perhaps you need to look > > through your other Postgres processes. > > Sorry, hang is the wrong word. Block is more correct. I know it's > waiting for commands from the front end, but I'm trying to figure out > what front end/query/lib/page is causing the problem. It's only once > in few thousand queries that it happens and I'm going crazy trying to > figure out what it is that's causing this. I've set > stats_command_string to true and attached gdb to the backend, but > that's null. Is there another var in the backend that contains a > pointer to the last query performed? I haven't found it if there > is... -sc Hrm, guess this'll work: log_pid = true log_statement = true log_duration = true log_timestamp = true Lower tech, but seems to work well effective. I changed pg_ctl to zip my error logs though, otherwise it grows too quickly. :-/ - "$po_path" ${1+"$@"} ${PGDATAOPTS+$PGDATAOPTS} </dev/null >>$logfile 2>& 1 & + "$po_path" ${1+"$@"} ${PGDATAOPTS+$PGDATAOPTS} </dev/null | $gzip -${ziplevel} -c >>$logfile 2>& 1 & -sc -- Sean Chittenden
В списке pgsql-general по дате отправления: