Last night two Apache threads hung using ~100% CPU and needing to be KILLed:
root 20139 0.0 0.6 221440 28140 ? Ss Mar31 3:19 /usr/sbin/apache2 -k start
www-data 31091 99.5 0.4 265340 19172 ? Sl Apr24 781:38 \_ /usr/sbin/apache2 -k start
www-data 31404 99.6 0.5 337932 21660 ? Sl Apr24 780:24 \_ /usr/sbin/apache2 -k start
In the Apache error.log are the following mod_pagespeed errors:
[Wed Apr 24 22:47:25 2013] [error] [mod_pagespeed 1.5.27.1-2857 @32668] Serf status 110(Connection timed out) polling for 3 threaded fetches for 0.05 seconds
[Wed Apr 24 22:47:25 2013] [error] [mod_pagespeed 1.5.27.1-2857 @32668] Serf status 110(Connection timed out) polling for 2 threaded fetches for 0.05 seconds
Munin graphs confirm a spike in "CPU usage" and "Interrupts" starting at the same timestamp.
We are running two instances of Apache with a total of up to 100 children. Other threads were unaffected.