Have you considered doing a daily reboot?

In our shop (about 14 sites, busiest doing about 90 000 calls per day) we found it best to reboot each Asterisk instance at 23:45 – we’re still running an ancient version (1.8.0.11).

Once this regime was instituted our major Asterisk issues stopped, especially load-related issues linked to uptime.

Note that we do not just restart the Asterisk process, we reboot the entire machine. We’re running on bare-metal on CentOS 6.5.

Maybe you can try doing that, if at all practical? Our business needs fortunately this side is telecoms are only requires 08:00 – 17:00…

Regards

Stefan

Message: 2
Date: Thu, 5 Feb 2015 11:53:38 +0100
From: Sebastian Damm
To: Asterisk
Subject: [asterisk-users] constantly increasing load in Asterisk 11.14
Message-ID:

Content-Type: text/plain; charset=”utf-8″

Hi,

we have quite a few Asterisk machines running and try to keep them on a current version of the Asterisk 11 branch. But since we upgraded to 11.14.0
a couple weeks ago, we have to restart the Asterisk process every week because the load gets too high and our monitoring complains.

Those machines are doing only SIP-to-SIP call relay, the dialplan is quite complex, transcoding is done only on a few percent of the calls processed. During the daytime, there are at max around 200 SIP channels (100 calls)
running at the same time. After one week, one machine has processed about
170k calls.

I have uploaded a comparison of cacti load graphs for one week of a machine running with 11.14.0 and one running with 11.6.0: http://pbrd.co/1v0SO3R

As you can see, after a restart, both machines have about the same load.