SIP Trunk error 503

[Sergey 1907]

I ask me to explain the following:
I have CUCM 12.5 Pub and Sub. There is CUCM 9.1 Pub only (for tests). Made SIP Trunk from 9.1 to 12.5 (only on Sub !!!) and from 12.5 to 9.1 (All Nodes is checked). Calls from 9.1 to 12.5 pass, from 12.5 to 9.1 no (503 error). I add to 9.1 in SIP Trunk 12.5 Pub in addition to Sub. Calls in both directions work, I remove Sub from the trunk on 9.1, everything continues to work. I remove Pub and add one Sub again - from 9.1 calls to 12.5 go, back side - no

CallManager service is running on all three servers. Is that how it should be?

I accidentally found out now: if the phone from which I am making a call is registered to SUB, then everything works in both directions, if it is registered to PUB, then it does not work in the direction of 9.1. Probably this is how everything should work )))

[Elliot Dierksen]

Why are you only having the trunk run on the sub? If you don't have the 'run on all nodes' box, the trunk will only run on the CM's that are listed in the CM group for the trunk. Run on all nodes overpowers that, and the call will come from whichever CM the calling device (aka phone) is registered with. Unless there is some really compelling reason to restricts which CM nodes run the trunk, I would suggest setting the trunk to run on all nodes on both sides. That does mean that the SIP trunk device will need any entry for each CM in the cluster on the other side.

[Roger Kallberg]

What you are seeing is the expected result of the described setup. The call would use the CPE CM where the device is registered to extend the call. If that IP is not present on the far end of the trunk the traffic will be rejected.

[Sergey 1907]

...