30% of phones go offline and then reconnect

I have an install with 2 systems and around 400 phones, all local to a pbx. The pbx has tremendous horsepower and shows no signs of being stressed. Once or twice a day 30-50% of the phones will failover to the secondary server and then 3-5 minutes later register back with the primary server. There is no rhyme or reason which phones will failover. Does FreePBX have a UDP session limit? The phones do have a failover SIP server and every 15s they send a SIP option packet to the primary server to see if it’s there. Perhaps this is causing issues?

There are no hard limits. You would need to look at the logs to see if any overload mitigation was in progress, although my first suspect would be the network.

Most likely it is a network and/or firewall problem that is closing the sessions. To mitigate this, I initially set many useful parameters on the phone (Grandstream) to keep the registering as

Register Expiration 5 minuts
Subscribe Expiration 60 minuts
Reregister before Expiration 240 seconds
Enable OPTIONS Keep Alive
OPTIONS Keep Alive Interval 30 seconds
OPTIONS Keep Alive Max Lost 3 packets
SIP Registration Failure Retry Wait Time 20

also reduced the custom sip headers due to the fragmentation of the udp performed by the firewall (Fortigate)

In the end I cut the bullet, as the firewall got in the way with “alg sip” systems, I switched to register extensions over TLS with TCP and SRTP and now I live happily

I looked through the logfiles and found no warning signs for another outage this morning. The phones are spread out amongst various buildings which are all connected on a single network so I wouldn’t think there are any NAT issues here. The odd thing is when a segment of phones go offline even phones on the same switch as the system will failover to the secondary server and then come back. It’s not like just a single building is dropping offline. It will also happen in the middle of the night so I wouldn’t think it’s a resource issue

This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.