[asterisk-dev] Machine lockups after Asterisk upgrade

[Patrick Staehlin]

Hi

Some 14 days ago I've sent a mail [1] to the users-list about a machine 
locking up after receiving a call from the sip-provider.

Since then we've upgraded Asterisk to 1.2.7.1 (on 07/17/2006) but that didn't 
make the lockups disapear. 

To give you an outline of the "problem":

We've upgraded an Asterisk 1.0.* installation to Asterisk 1.2.4 and later to 
1.2.7.1. Since this upgrade we have occasional (~twice a day) hard lockups of 
the machine running Asterisk. The server runs a Debian Sarge with a Kernel 
built from Debian-sources. We got Asterisk packages are from 
http://debian.peen.net/asterisk

The lockups occur right after the caller hangs up his phone and doesn't seem 
to be related to the overall call time nor the uptime of the server. We 
suspect there is an runaway asterisk-process that dominates the cpu (I 
thought that couldn't happen with linux, but there it is), as a memory fault 
could be ruled out by a memtest86.

After the lockup the server seems to have a running kernel since it responds 
to SysRq-requests (although there seems no processor time left for any other 
processes). Through the SysRq we can sync & remount the hard-drives an get a 
more or less gracefull reset of the machine.
After terminating all processes (SysRq + e) the server gets responsive again, 
supporting out theory about a runaway process.

We're using a dialplan generated by freepbx. Out setup has been described in 
[1] but I'd be glad to post further information on request.

[1] http://lists.digium.com/pipermail/asterisk-users/2006-April/146708.html

With kind regards & thanks in advance!
Patrick Staehlin