[asterisk-users] SIP Mass exodus

[Chris Bagnall]

On 13 Nov 2013, at 18:29, Mike Diehl <mdiehlenator at gmail.com> wrote:
> I've been seeing some strangeness lately on my 10.2.1 server.  It's
> gotten to the point that a few times each day, I see masses of SIP
> clients becoming unreachable.  They're not all on the same network,
> and we don't see any calls drop.  In a few seconds, they all come
> back.
> I don't think it's a connectivity issue because we don't drop calls,
> and the endpoints aren't on the same networks.  We don't see excessive
> CPU load when it happens.
> It does SEEM to happen most right after someone accesses their voicemail.


We saw this happen on a 1.4 server a couple of years ago shortly after 2am each day. It was only after a study of the cron schedule we narrowed it down to a number of rsync backup jobs which were run at that time.

As in your case, it wasn't a connectivity or bandwidth issue - in the end we put it down to a disk I/O bottleneck. It might be worth running something like iostat on your box to see if you see a spike in iowait as voicemail is being checked. We resolved it simply by rate limiting our rsync jobs. In your case with a busy database, you might want to look at your MySQL indexes and/or cache settings - this might be something worth asking about on the respective MySQL discussion groups as well as here.

Kind regards,

Chris
-- 
This email is made from 100% recycled electrons