The httpd root process has been dieing seemingly at random on the
different app0* servers. We're not sure what it is, but there's a
chance that it could be the fedmsg-emit.php plugin. Here's a proposed
patch that's been sitting in staging for about a week:
The theory is that, before, if "fedmsg-config" failed (which it
shouldn't), then the script would call die(). die() shouldn't
terminate the root httpd process, but.. well.. we're not sure
what else could be happening.
The change goes: instead of calling die(), if the plugin can't
setup its fedmsg socket, then it just won't register its callbacks
with mediawiki, instead of calling die();
I've tested that this is sane in staging, but there's no way to test
that it fixes the issue, since it happens randomishly.
This doesn't require a new fedmsg rpm. The fedmsg-emit.php plugin
is kept in puppet.
Seeking +1s for freeze break.