[Asterisk-Users] Asterisk crashes from time to time

[Rich Adamson]

Inline...

> > Far too many variables for anyone to even guess at the root cause. Problem
> > could be related to slight differences in o/s libraries between systems,
> > coding problems within asterisk, etc.
> You 're right, it could be every thing....
> 
> > There were some issues reported with cvs head in January relative to
> hangs,
> > etc.
> Are they reported in the bugtracker, or in the mailing list?

Not all for sure. If you watch the -cvs and -user list, you'd see folks
with seg faults, etc, and not too long after that you see a change
come through -cvs. Sometimes with comments like 'fix silly typo', etc.
Given that cvs head is actually development, at any point in time there
could easily be various problems (expected). To try to recreate
historically whether you caught a cvs head version that had errors is
almost impossible. 

That's why its important to run cvs head in some sort of pre-production
test environment before promoting the code into a customer's machine, etc.
(That implies beating the hell out of your test environment.)

> > Might consider changing /etc/asterisk/logger.conf and add debug to the
> list.
> > Then after a failure, at least look at /var/log/asterisk/debug messages.
> Yes, this was the first thing, I did after the crash showed up. I simply
> forgot to enable it, since this production server ran long time without
> problems. But now, following murphy's law, the next crash will never happen
> ;-)
> 
> > For additional info, I'd suggest compiling the code on one of thse
> machines
> > to see if it complains about missing/inappropriate items.
> 
> After these machines were setup, we compiled new code on every machine,
> since we started with an older version of Asterisk in November 2004. The
> compiling of asterisk did not show me any relevant (?) errors. But I
> remember there were some statements (Warnings) in the console output of the
> make process, I didn't understand. Is this output logged in addition to the
> console in a logfile somewhere?
> If so, one could examine this output and hopefully get some hints...

The only two (key) log methods that I know of is to run the cli with
several -vvvvvvvvvvvv's, and turn on debugging in the logger.conf 
file (which may require you to config /etc/syslog.conf to catch them).
Then look at /var/log/asterisk/debug after a failure.

(There are other debug modes, but not sure I'd use those to catch a
production problem. The one's I know about are primarily intended for
development debugging. Other folks might contribute hints here.)