On 08/10/2012 10:06 AM, Tom Lane wrote:
> Greg Williamson <gwilliamson39@yahoo.com> writes:
>> Every minute on the minute there is a cronjob that kills IDLE IN CONNECTION procs older than 3 seconds ... possibly
thatprocess killed the process I was running, but to me the user it seemed as if the COMMIT came before the disconnect.
> That sure sounds like the source of your problem. It might be sane if
> it killed only processes that *had been idle* for at least three
> seconds, but I'm not sure there is any easy way to determine that ...
wouldn't:
select * from pg_stat_activity
where current_query = '<IDLE> in transaction'
AND query_start < current_timestamp - INTERVAL '3 seconds';
do it?
--
Craig Ringer
