httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jeff Trawick <>
Subject [PATCH] tweak worker MPM "recovery" after pthread_create failure in child
Date Thu, 05 Aug 2004 12:29:23 GMT
Instead of killing the server, let it continue on.
Instead of sleeping in the child, exit ASAP from child to possibly
free up resources and do the sleeping in the parent.
Any comments/concerns?  (Patch is against 2.0.50 unfortunately; I
can''t get HEAD of either branch to build at the moment.)

I've been working with a user on RedHat {A|E}S 2.1 that is
intermittently getting one of these thread create failures as the web
server tries to create a new child process to handle increased load. 
It really sucks that the entire server bails out.  Knowing the root
cause is not so easy.  Monitoring of process memory use via ps and
overall system memory use via /proc/meminfo never shows any situation
where any processes are using lots of memory or where any significant
amount of paging space is used.  The thread create failure wouldn't
seem to be due to exceeding a process or user limit since the server
is able to create enough threads to reach MaxClients at some observed
times.  I'll probably give them a diagnostic build which logs
/proc/meminfo and ps listing at the time of the thread create failure
to see if there was a temporary spike.

View raw message