On Fri, 7 Sep 2012, John Hardin wrote:
On Fri, 7 Sep 2012, Kevin A. McGrail wrote:
On 9/7/2012 11:23 AM, John Hardin wrote:
> > My best guess is the lag in response could be causing SA to spawn
> > more children which could be spiraling memory to swap, etc.
>
> They contacted me directly off-list and we discussed swap. They said at
> one point it _was_ hitting swap, but they reduced setting and it is no
> longer doing so, but it is still showing very high CPU usage.
>
> I will be talking with them again this morning.
Would DProf be a good idea to enable for profiling?
Possibly, I don't know that we're to that point yet.
At this point it _seems_ to be due to something in Amavis. They are
running a fairly old release (2.6.4), and the observed behavior is that
all of the Amavis daemon child processes are running as fast as they can
(hence the 100% CPU across 4 cores and a load average of >20) regardless
of whether or not any messages are being processed, and the daemons ignore
the shutdown signal when the service is stopped so restarting Amavis
leaves the processes running at 100% CPU.
Shutting down Amavis, kill -9'ing all the amavisd child processes, and
restarting Amavis seems to correct the problem, but how long that will
last is not known yet.
I have recommended that they consider upgrading to the latest Amavis, as
2.6.4 is 3 years old now.
--
John Hardin KA7OHZ http://www.impsec.org/~jhardin/
[email protected] FALaholic #11174 pgpk -a [email protected]
key: 0xB8732E79 -- 2D8C 34F4 6411 F507 136C AF76 D822 E6E6 B873 2E79
-----------------------------------------------------------------------
10 days until the 225th anniversary of the signing of the U.S. Constitution
