For the second time in a looong time I have the problem that only 50% of the calls from a GW to a SIP-phone come through, the other 50% go to voicemail directly. I have an HA setup and the GW is distributing the calls round-robin to the 2 SipX servers that form an HA-cluster. Also calls from phones registered on one server (S-A) of the HA cluster to a phone with the problem on the other server (S-B) will go to voice mail directly. I am using 4.2.1-018971 and Bria 3.1.2.1
Most phones (Bria's) work OK, but some show the following behaviour: When the call flows to the SipX server (S-A) that holds the registration for that phone then the phone starts ringing. When the call flows to the other SipX server (S-B) then you go to voicemail directly. I have traced a working Bria and one with the problem. The difference lies in the port that is being used to send the invite on. BTW: my phones are configured to use TCP. In the normal case the phone registers from port A to 5060 (syn, syn ack etc) on S-A. The phone then sends some Notify requests over UDP from port B to 5060 on S-A. If the call flows through S-A (the server that holds the registration) then the existing session (5060 to port A) is used. If the call flows through S-B then the server sends a SYN from a random port to the phone to port B (so the same port that the phone used to communicate with S-A for UDP communication) This is SYN-acked and all goes well. In the case of the problem all is the same for S-A and these 50% of the calls work. If the call flows through S-B however then the SYN is not send to port B (the UDP port used by the phone between phone and S-A), but to port A (the TCP port of S-A). This SYN is generously RST-ed and you end up in voicemail. I found a really old issue, http://track.sipfoundry.org/browse/XCL-89 maybe something like this can happen very rarely as well. If needed I can send wiresharks, snapshots or anything else, please let me know. Restarting services does not help. Resetting the servers one at a time does not solve the problem as well. The only remedy so far that sort of works is resetting both servers at the same time. Paul
_______________________________________________ sipx-users mailing list sipx-users@list.sipfoundry.org List Archive: http://list.sipfoundry.org/archive/sipx-users/
