On Thu, 10 Jul 2008 12:57:16 +0100 "Benny Prijono" <bennylp at pjsip.org> wrote: > On Thu, Jul 10, 2008 at 2:42 PM, Will Davies <will at hugthebubble.co.uk> > wrote: > > > > That should be okay, that STUN attribute is optional. Are you sure it's > > > pjsip that's dropping the call and not Gizmo? > > > > > thanks for the response > > That would seem unlikely, how would I tell? > > users A, B, C and D all have gizmo accounts > > users A and B can hold a conversation for over 40 minutes > > any call involving users C or D generally dies after 6-10 minutes (maximum > > so far 12) > > users A and B have static internet addresses > > users C and D have dynamic internet addresses > > Its not much as far as proof goes. I'm slightly overwhelmed by the > > verbosity of the log and in over my head. Any pointers on things to look for > > would be gratefully received. > > > Yeah I know what you mean about the log. But look for the part where BYE > message is sent or received, and see who sends it. If the sender is pjsip, > scan few lines before that, normally it should print out why it hangs up the > call (unless of course if it's user who hangs up the call). There is no BYE until one of us hangs up. I have a log that include several minutes after the sound stopped. The only think I have managed to learn from it is that the 18:37:45.594 strm0x823ce04 Starting silence 18:37:45.595 strm0x823ce04 Start talksprut.. messages on the machine with the high cpu utilisation stop at the same time that the call stops working. I continue to get these messages at the other end but neither of us can hear anything > > > The other point to note is that users C and D see very high cpu utilisation > > when the call dies. > > > > Okay that's interesting. If the CPU utilization is very very high, then it > could cause some SIP transactions to timeout with 408 status, and this > status will terminate the call indeed. But I don't think this would happen > unless the CPU utilization is very very high (to the point that the computer > feels like it's locked up). > No 408s in the log thanks Will Davies
