Expressway C&E clustering issue

[Attila Horvath]

Hi, 

We had setup an Expressway-E cluster with Expressway-C cluster.

The system is NAT-ed, so the Edge pair is with Dual interface (Cluster peer interface is not the NAT interface -  according Guide).

The DNS srv is load balancing between two Edge peer (same priority).

When the cluster is UP and running, the Jabber endpoints close the session after a short time (some minutes), also the registering to the CUCM is sometimes unsuccessful .

When the slave peer of  edge cluster has switched to Maintenance mode, everything works like a charm. (The H323 is on at peers.)

What could be the problem?

[Elias Sevilla Duarte]

Hello,

This sounds like a connection issue in the traversal zone between the Core and Edge cluster peers.

When the system is in maintenance mode, it works like a charm. This could be because when the peer that is in maintenance mode is contacted, it replies with a 503 Service Unavailable, so Jabber does a new request for registration, and it goes to the peer that is active, and the SIP registration flow always uses the same path. When both peers are up, it may use either of them for the traffic flow, and if one of the links is down, the registration will fail until a new request is made, resolving to the peer that has no connectivity issues. I have seen this before, so I do recommend to check the traversal zone when the issue takes place. In any case, the answer to this could be captured in logs if taken realtime.

HTH.

[Attila Horvath]

Hi, 

We have checked several times, and all the SSH tunnels show UP between C and E.

Also note, if I replace the Maintenanced E host  (change the Maintenance status to the pair of E cluster), it is also works.

So it is independent of E host.

[Elias Sevilla Duarte]

Hello,

Perhaps you can download logs and use the Collaboration Solutions Analyzer to verify if there is any specific error that you can be pointed to.

https://cway.cisco.com/tools/CollaborationSolutionsAnalyzer/

HTH.

[Attila Horvath]

Hi, 

Thanks for link, I checked MRA, and it shows everything is OK.

Also tried to check Core/edge and Jabber log files - but not found any interesting issue.