httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Graham Leggett <minf...@sharp.fm>
Subject Re: sigalrm and such
Date Wed, 02 Jun 1999 10:05:38 GMT
Dean Gaudet wrote:

> Graham, this could be related to your unkilled child problem... give the
> patch a try.

I pulled down this morning's CVS tar.gz of v1.3.7-dev, which has the
patch already (as far as I can see).

The patch has no effect.

If you look at the Apache scoreboard file by looking at
http://infobase.ericsson.se/server-status you'll see a whole lot of
processes are stuck in a graceful restart state. Apache is gracefully
restarted once an hour on the hour to do logfile rotation. Over time
most of the processes end up stuck in the graceful restart state.

If you send a apachectl restart request the logfile indicates that the
processes involved ignored both a SIGUSR1 and a SIGTERM, and all the
processes are SIGKILLed.

> I dunno, I always just suspect it's someone doing an alarm() or sleep()
> somewhere.  But maybe that "someone" is a stupid solaris library bug (a
> few have been known to exist)... and not a 3rd party developer.

I would strongly suspect Solaris v2.6, because our problems grew worse
when we downgraded from Solaris v2.7 to v2.6. In our case v2.7 panic'ed
all the time, now we contend with v2.6 leaking all the time. Sigh.

Any further ideas?

Regards,
Graham
-- 
-----------------------------------------
minfrin@sharp.fm		"There's a moon
					over Bourbon Street
						tonight...

Mime
View raw message