I am running Asterisk 11 on CentOS 6.x using the DAHDI module with 8x PSTN analog phone lines for outside connectivity. Internally, I am using several models of Yealink SIP phones (e.g SIP-T32G) on a dedicated VoIP network, 192.168.0.0/24. I have a few of these Yealink SIP phones configured with an OpenVPN certificate so that users working remotely can directly access the phone system (VPN subnet is 192.168.1.0/24). Note that this is not a NAT; VPN clients are able to directly address the Asterisk server and other SIP phones. Last week the phones connecting over the VPN started dropping audio during the call (e.g caller 1 can still hear caller 2, but not vise versa). These calls are between two SIP phones (one over the VPN, one internal). The dropouts last for 20 seconds or more, and sometimes the audio does recover and come back.

I made some changes to the infrastructure last week, but I am not sure that they are the cause. First, I added echotraining=yes to /etc/asterisk/chan_dahdi.conf to try and fix echo problem (seems unrelated since the call is all SIP). I also cleaned up some extraneous firewall rules on the OpenVPN gateway, but I still allow the VPN phones to connect to the Asterisk server on ports 5000 – 20000 for SIP and RSTP so this also seems unrelated.

I’ve looked at the syslog on the SIP phones as well as the asterisk output with “sip set debug” and “rtp set debug” on but I don’t see anything obviously wrong. The only sign of a problem I can see is this message when the call is hung up:

pbx.c: == Spawn extension (dial-extension, 124, 1) exited non-zero on ‘SIP/123-000001d9’

Here is an example user in my sip.conf:
http://pastebin.com/6U2AhyWT

Do you have any ideas about what is causing these dropouts, or what I should look at next for additional debug information?