Packet loss with increasing call load

logging is not an issue unless your storage is really slow (like post office slow). we have run 75 concurrent calls on a dell r610 and it barely keeps the machine warm. if i remember correctly amazon does not use off the shelf systems, but builds their own supposedly to be more efficient but i doubt their architecture would impact freepbx. if it were me, i would spin up a new system on the dedicated machine and then restore it from the running one. to make life easy for yourself , if amazon will allow it, spin up the new system and take a snapshot before doing the restore. that way if the restore craps out or exhibits the same behavior you can quickly revert to the snapshot and manually rebuild your system.

‘Unable to write frametype 2’ is an OS level issue. Asterisk is trying to write to a socket, and the socket isn’t working. Restarting asterisk should fix it, but the actual cause is in the host or the hypervisor.

To add my two bits, I am running a virtualized FreePBX, BUT on MY servers. It runs low use processed other than the PBX. File servers, DB and Email are on a separate host. I’ve done this because I don’t trust public clouds to give me the Juice I need for the voice.

1 Like

As an update, it seems the packet loss was mostly fixed by replacing all of my edge switches, and the issue with Asterisk dropping everyone’s calls at once was actually an Asterisk bug with PJSIP; it hasn’t been resolved yet but rebooting each night seems to be preventing the system from crashing completely. I was posting in the Asterisk thread for the issue here:

1 Like