Running Edge 2024.6.1 on Dell R750 nodes here in Tokyo. BIOS sits in performance mode with C-states disabled to keep the CPU steady. Primary WAN link dropped yesterday during ISP maintenance. Edge automatically triggered survivability mode. Local SIP phones registered fine through the secondary NIC. The on-prem SIP trunk keeps bouncing though. Asterisk logs show a steady stream of 487 Request Terminated messages every thirty seconds.
Local media bridge looks stuck on the RTCP packets. Checked the interface metrics on eth1. Packet loss sits at zero. Bandwidth is plenty. SIP OPTIONS probes from the carrier side return 200 OK initially, then drop after the third cycle.
We’ve already bounced the SIP stack service, but it’s doing jack all. Tried tweaking the timer settings in the Edge portal. Set invite_timeout to 45s. Didn’t help. The trunk just keeps resetting. We can’t get the carrier side to hold the dialog open. Network team verified the MTU size matches the upstream router at 1500. Jumbo frames are off.
Edge logs show this right before the drop:
[2024-11-15 09:14:22.881] WARN o.g.c.e.s.sip.SipStack - Transaction timer F expired, sending CANCEL
[2024-11-15 09:14:22.905] ERROR o.g.c.e.s.sip.SipStack - SIP-12004: Dialog terminated unexpectedly (487)
Checking the kernel ring buffer now to see if the NIC driver is throttling the TX queue.