[Freeipa-devel] apache segfaults

[Alexander Bokovoy]

On Fri, 09 Nov 2012, John Dennis wrote:
>I'm wondering if anyone else has seen this. I've been running the 
>server with debug=True to verify it's behaving properly which means 
>I've been reading /var/log/httpd/error_log and what to my wondering 
>eyes did appear?
>
>[core:notice] [pid 9089] AH00052: child pid 9092 exit signal 
>Segmentation fault (11)
>
>The server seems to be running normally, except after the above 
>segfault notice there are debug messages about loading our plugins 
>which seems to imply it's restarting a wsgi process, but it's not 
>clear that's what's actually happening. This is on a F18 vm.
>
>The only other information in the log is:
>
>*** glibc detected *** /usr/sbin/httpd: corrupted double-linked list: 
>0x00007f60952cddc0 ***
>
>So a few questions:
>
>* anybody seen something similar?
I do see it, on F18 VMs.

>
>* does apache respawn a child process that crashes? (the log seems to 
>suggest so)
yes.

>
>* does systemd respawn a daemon that crashes?
Yes, if it is the main process for the service from systemd perspective.

No for Apache since master process did not fail from perspective of
systemd.

>
>* how does one get a coredump? (Is that an apache config item?, a 
>systemd flag?)
Theoretically, systemd sends all coredumps it sees to journal. systemd-coredumpctl(1) is
then used to list and extract coredumps.

>* any reason to believe this has anything to do with us?
I vaguely remember seeing some bugs in the Fedora Bugzilla about
mod_wsgi failures in recent Apache updates on F18. Cannot find them now.

>* any suggestions for diagnosing this?
Try to set LimitCORE in httpd.service to non-zero value (as in
RLIMIT_CORE from setrlimit(2)).
Reload systemd configuration with 'systemctl daemon-reload' and hope to
get core in the journal.


-- 
/ Alexander Bokovoy