cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
688
Views
0
Helpful
3
Replies

Unity Connection upgrade - Cluster Requirements

RossRoach
Level 1
Level 1

Hi All,

I need to upgrade our Unity Connection cluster from version 7.0(2) to a later version (looking at 7.1(3b)SU2).  I've noticed that in the "System Requirements for 7.x" under "Cluster Requirements When the Servers Are in Separate Buildings or Sites" it mentions that both servers need to belong to the same platform overlay.

Currently we have two servers in our cluster (located at separate sites), one being a 7845H2 and the other a 7835H2 (Platform overlays 3 and 2 respectively).  Does anyone know what issues we should expect to see occur as a result of having two servers in different platform overlay groups?

Our system was installed by a 3rd party, and this is the first time i've had to look into the clustering requirements.

We have had TAC calls open in the past looking into failover issues with our cluster, and none have mentioned this incompatibility, only suggested that an upgrade may help.

Any replies would be appriciated!

Thank you.

1 Accepted Solution

Accepted Solutions

David Hailey
VIP Alumni
VIP Alumni

I suspect the main reason you're being told you should upgrade is because you're on 7.0 which was riddled with bugs.  7.1(3) and specifically 7.1(3b)SU2 (my recommendation to you) has been very stable.  You are correct about the requirements to use the same platform overlay.  From a "will it work if I don't" perspective, the answer may likely be no.  I've never looked that deeply into the requirement because, personally, I have not installed nor would I recommend a mixed cluster from a hardware perspective.  The reasoning being that Unity Connection is intended to provide 100% failover between servers. So, if you have a 7845 and a 7835 then the effective capacity of your cluster is that of a 7835.  This means you have to be cognizant of what load is being placed on the servers and that you distribute workload and overall cluster size (e.g., subscribers, ports) such that, in a failover scenario, you don't overload the smaller server.  This requirement actually stems back to the original Cisco Unity.  Some may not know but, in a Unity failover scenario, you should match the 2 servers in overlay as well.  Some of this is due to file replication and such but again some of it is due to server sizing and have equal capabilities between servers during failover.  With CUC, the architecture at a high-level is similar to that of CUCM and while you can have a mixed cluster in CUCM (different server types) - best practice recommendation is to use the same platform across the cluster.

Hailey

Please rate helpful posts!

View solution in original post

3 Replies 3

David Hailey
VIP Alumni
VIP Alumni

I suspect the main reason you're being told you should upgrade is because you're on 7.0 which was riddled with bugs.  7.1(3) and specifically 7.1(3b)SU2 (my recommendation to you) has been very stable.  You are correct about the requirements to use the same platform overlay.  From a "will it work if I don't" perspective, the answer may likely be no.  I've never looked that deeply into the requirement because, personally, I have not installed nor would I recommend a mixed cluster from a hardware perspective.  The reasoning being that Unity Connection is intended to provide 100% failover between servers. So, if you have a 7845 and a 7835 then the effective capacity of your cluster is that of a 7835.  This means you have to be cognizant of what load is being placed on the servers and that you distribute workload and overall cluster size (e.g., subscribers, ports) such that, in a failover scenario, you don't overload the smaller server.  This requirement actually stems back to the original Cisco Unity.  Some may not know but, in a Unity failover scenario, you should match the 2 servers in overlay as well.  Some of this is due to file replication and such but again some of it is due to server sizing and have equal capabilities between servers during failover.  With CUC, the architecture at a high-level is similar to that of CUCM and while you can have a mixed cluster in CUCM (different server types) - best practice recommendation is to use the same platform across the cluster.

Hailey

Please rate helpful posts!

Thanks for your reply Hailey.

Unfortunately I can't see us replacing either of the servers for a while, so I haven't really got a lot of choice but to give it a shot and hope for the best!  Both of our Unity servers are well underutilised, so although I'll have to be mindful of keeping within the limits of the 7835 it shouldn't be too much of an issue from that respect.

Thanks again, much appreciated.


Ross.

Yes, different types of servers in the same cluster for UC is unsupported.  The active-active cluster design requires that both servers hold all the messages and are capable of taking 100% of the calls so it only makes sense to require they be basically the same server.

If you have issues that TAC believes are a result of having two different servers in the same cluster TAC will just end the converstation with sorry but what you are doing is not correct and you need to do it the supported way before we can go any further.

You won't likely hear anything about it being unsupported unless the problem is directly related to something we believe might be caused by this mis-match.

So where to go from here.  Try to get the same server.  Maybe your account team can help swap one of them out.

If thats not possible limit the usage to the smaller servers level so you don't stress the smaller one.  It's going to be setting your limits on how many ports, how much voicemail you can save.  If you go over those limits there is no safety, no way to know what will happen and TAC will tell you to turn one of them off and remove it from the cluster.