[asterisk-users] What do you do to keep asterisk alive?

[Luki]

> We tend to find that on the rare occasions asterisk does decide to die,
> it very often doesn't die completely.

Agreed. Need to also watch out for SIP deadlocks (asterisk is up, you
can connect to the CLI, but it does not respond to any SIP traffic, or
sip reload, or unload chan_sip, or restart). Sipsak via an external
monitoring script works for us. On a timeout it first tries to
gracefully stop asterisk, then it force kills it with kill -9, and the
restarts it. That has proven successful in minimizing downtime.

On the same machine, same binary, "same" traffic volume, sometime
asterisk stays up for months (current record, 51 weeks!) and sometimes
it will decide to die after couple days (current record, 6 times on
the same day). A core dump doesn't reveal anything suspicious. It's
not load related as crashed can happen at 3am when it's pretty quiet.
I suspect it's due to network issues (dropped packets, etc) because
sometimes just before the crash the console is full with
__sip_destroy: Trying to destroy ... not found in dialog list?!?!
messages. No, I have not upgraded to 1.4 yet.

Still, I find it ironic that all this effort goes into fixing the
symptom rather than the cause. But I'm not complaining... just don't
have a better idea how to fix it.

Luki