On Wed, 2010-10-20 at 16:26 -0400, Tom Lane wrote:
> > It seems strange that the postmaster is eating 99% cpu. Is there a
> > chance that it's flooded with connection attempts?
Maybe, I'll try to figure that one out next time it happens.
> It's probably a backend process, not the postmaster --- I suspect the
> OP is using a version of ps that only tells you the original process
> name by default. "ps auxww" or "ps -ef" (depending on platform)
> is likely to be more informative. Looking into pg_stat_activity,
> even more so.
I'm running CentOS 5.5, using procps-3.2.7-16.el5. I cannot check
more at this point as postmaster seems to have finished whatever it
was doing, but I'll try to investigate better next time.
--
Dimi Paun <dimi@lattica.com>
Lattica, Inc.