07-19-2017 08:12 AM - edited 03-18-2019 01:19 PM
I browsed the CMS documentation but could not find any information regarding conference and call behaviour on a CMS Callbridge cluster when one of the members is failing.
So suppose I've a CMS cluster of 2 nodes: CMS1 & CMS2. And a 3rd CMS just for the DB clustering.
I have a space that is active on CMS1 with users connected.
Now CMS1 drops due to a network or HW issue.
1. Will the active space be recreated on CMS2?
2. Will all users be disconnected from the space?
3. Will all users need to dial the space ID again to rejoin the space?
4. What happens with users connected via WebRTC?
07-19-2017 08:44 AM
I will know more by next week as we will be doing full fail-over testing end of week but from what I've tested so far there isn't really any fail-over perse. If someone is homed on CMS1 and it fails they get disconnected. If they are on CMS2 they stay in the call. For the users who were on CMS1 and got disconnected, when you call back it will work fine as it will take you to CMS2. One extra variable in this though is the XMPP server part. For now, the XMPP resilience is still in beta so if the CMS that fails has the XMPP homed on it then any client that depends on XMPP will lose connectivity and won't be able to reconnect until XMPP is back.
Again, I'll be going through a bunch of explicit scenarios (we have two Ccore plus a DB server as well as two Edge servers) so will know more then
Steve
07-19-2017 11:00 AM
If the connected endpoints and WebRTC participants were on CMS1 when it failed or dropped from the network, they would need to redial into the conference, which depending on your dial plan would put them on CMS2. Anyone connected to CMS2 when CMS1 went down, they would be unaffected, and it would appear to them that anyone on CMS1 simply disconnected from the conference because the distributed link between the call bridges would be broken.
Stephen mentions XMPP resilience is in beta, since CMS 2.1 it has been a fully released feature, no longer in beta, and supported in production environments.
07-19-2017 11:06 AM
good to know the XMPP resiliency is out of beta. Do you know what licensing is needed to make that work if we already have an XMPP license? Since we have two Core servers plus a DB server (which doesn't require a license) I'm guessing we'd need to license that "DB Server" if we wanted it to be the third XMPP server? Or if it is part of a cluster it just has one license?
07-19-2017 11:34 AM
XMPP requires an activated Call Bridge.
07-19-2017 11:43 AM
weird that a CallBridge is require (which they mention in the Deployment guides early on) but on page 69 of the "Scalable and Resilient..." Deployment Guide (I have the 2.1 version) it showsa diagram of having 3 XMPP servers (for a quorum of a cluster) and only 2 CallBridges which led me to believe you can run it as a stand alone. If not then I guess it kills the whole idea of having your 3rd server be just a DB server. I guess I'll need to verify if that is still the requirement for an XMPP server now that they've added the resiliency.
Steve
07-19-2017 12:00 PM
I agree, I have a similar deployment as you: 2 CMS + 1 Database. All I can say is in every deployment guide, regardless of deployment model:
The XMPP license is included in the Cisco Meeting Server software. You will also need a Call Bridge activated on the same Meeting Server as the XMPP server.
Discover and save your favorite ideas. Come back to expert answers, step-by-step guides, recent topics, and more.
New here? Get started with these tips. How to use Community New member guide