[asterisk-users] Crash Hard, Crash Often

[Josiah Bryan]

Paul Chambers wrote:
> Josiah Bryan wrote:
>> <snip>
>> Problem is that its crashing for seemingly no reason at all, no errors 
>> on the console, no logs (that I can find), nothing in /var/lib/messages 
>> - its puzzeling! Management is screaming like banshees, calls are 
>> dropping like flies, and all hell is about to break loose if I can't 
>> stop asterisk from crashing every couple of hours, taking down any 
>> Zaptel calls with it.
>> </snip>

<snip>
> That description reminds me of a problem I ran into a while back. One 
> fan had quietly failed, and the temperature would slowly creep up inside 
> the box until things started 'acting funny' and the box would lock up 
> soon after. It'd run fine for 3-4 hours, then just keel over and die. 
> The logs didn't show anything consistent just before the event.

The wierd thing is that its *just* the asterisk process that dies - the 
rest of the system stays solidly up...

<snip>
> Do you have another PC you can swap the drive and cards into, to try to 
> rule out hardware instability? could you run lm_sensors? (along with one 
> of the logging/alarm packages that support it).

Well, Paul, it looks like that was indeed the problem (hardware 
instability.) I came into the office last night after everyone left in 
order to swap out the RAM in the server - lo and behold, I didn't have 
any of that type of RAM around (RIMM's ??), so I had to do an emergency 
hard drive & PCI card transplant to a similar chassis.

After a bit of tweaking to get ALSA to work right and the NIC to play 
nice in the new chassis, asterisk came online and worked beautifully. 
(And, shockingly enough, the zaptel cards just *worked* - no tweaking 
needed!)

So far, no crashes today (by this time, normally it's crashed two or 
three times already in a day.)

So, we'll see how she runs - If I were a betting man, I'd say that 
something in that old chassis was going out - probably the RAM as stated 
before, but not sure.

As far as the power supply being "good", I believe it was - didn't 
check. The server was a re-purposed high-end CAD workstation - the 
dismal RAM and CPU belie the solid construction of the chassis and the 
quality of the workmanship in the way the server was put together.

Now that I've waxed weird, I'll just say the hardware seems to have been 
the problem and I'll keep and eye on it. This may have yet saved me from 
converting over to the callweaver fork - we'll see. :-)

Cheers!
-josiah

-- 
Josiah Bryan
IT Manager
Productive Concepts, Inc.
jbryan at productiveconcepts.com
(765) 964-6009, ext. 224