Hi Jano,

> BUT, seems like apache does segfault if graceful restart is done:

Graceful restarts always have been a problem with apache, even with the
normal prefork MPM. I've never seen segfaults there but processes
hanging forever are not that uncommon.

The code in peruser handling this situation is somewhat complex and to
be honest I'm glad I didn't have to touch it so far. Sean, is this part
specific to peruser or did we inherit that code from some other MPM?

I cannot reproduce this and unlike the previous problem and don't even
have an educated guess. My best bet was a race on the pipe_of_death but
I can't get a segfault even when I run an apache bench with 20
concurrent requests against the server.

> About 7-8 of segfaults seem to happen after (few seconds or less)
> graceful restart. Probably all processors which are active at the moment
> seem to crash and burn.

o Can you please check the log files and see if those processes have
been active (serving requests) before the graceful restart or if they
were just starting?

o Is the server a multi processor machine?

o Do these Segfaults always happen? If not

 - Does it matter whether the config has been changed or not?
 - Does it depend on the load of the server (e.g. only happens if the
   server is heavily loaded)?

o Do these segfaults only happen with a graceful restart (SIGUSR1) or do
they happen with a normal restart (SIGHUP), too?

o Do you send the signal only to the master or to all processes (kill
vs. killall)?

> This does not seem to be big problem, but if someone feels bored then it
> could be fixed and if not fixed as a feature , it could be fixed as a
> little cleaner notice than segfault.

You are too kind but I think segfaults always are problems which should
be fixed ASAP.

Thanks,

Stefan

_______________________________________________
Peruser mailing list
[email protected]
http://www.telana.com/mailman/listinfo/peruser

Reply via email to