UCCX 5.0(2) Questions

Unanswered Question
Apr 26th, 2010

Hello,

I'm running UCCX 5.0(2) in an HA configuration.   I have two questions.   My first is sometime over the weekend, my secondary server became the master for most of my services but not all.   I cannot find what caused this failover and do not understand why only some of them failed.   Any ideas where to look?

My second quesiton is, some of my users are using the single button phone login.   When my servers failover, my users cannot use the single button phone login service.   I know that I can create another single button phone login service that points to the secondary server but I would prefer not to do this if there is a way around it.   Is there a way I can tell the primary server to forward the login request to the secondary server when the primary is not the master?

Thanks,

Glenn

I have this problem too.
0 votes
  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 0 (0 ratings)
Loading.
Aaron Harrison Mon, 04/26/2010 - 14:48

Hi

HA - Basically what I guess you are seeing is that the DB services are Master on the primary/publisher node, and that the 'Desktop' and 'CCX Engine' services are Master on the secondary/subscriber node. This is normal - basically if the publisher is up, it is always Master for the DB services - meaning the sub/sec server writes any changes back to the publisher.

It's only if the Master is fully down that all services go to the secondary. The important services are the desktop/CCX engine ones; if the server that these are running on experiences a problem you'll get a failover and drop calls. If the publisher is master for the SQL services and fails, it usually goes unnoticed.

Re: the single button login - you need to do one of two things:

1) Have two services as you describe

2) Have a clever enough load balancer to switch traffic to the other server if it recognizes a particular response from the service (i.e. the 'this service is not active' message).

There's not much point having server A forward requests to server B, as the idea is that if server A isn't there, you use server B. If server A isn't working, then chances are it can't forward your request. I guess in theory it would make sense for it to redirect if it is in a good enough state to give back the 'not active' message, but it's certainly not an option in current versions.

Regards

Aaron

Please rate helpful posts..

glenn.martindel... Tue, 04/27/2010 - 06:43

Aaron,

Thanks for the response.   I guess I really didn't mean that the primary would forward the request to the secondary.   I just stated it that way because it was shorter to describe.   It just seems odd to me that CAD will fail over to the secondary server if the primary is down put the phone login cannot.   Especially if the phone agent service is going to flip over to the secondary at will.  i can't use CAD because some of our users are running Windows 7 and our version of CAD won't install on Windows 7.

Could I initiate the "load balancing" through DNS?

Thanks,

Glenn

Aaron Harrison Tue, 04/27/2010 - 11:58

Hi Glenn

The problem with DNS (and I presume you are suggesting round robin DNS) are two really:

1) Cisco phones don't play nice with round robin DNS. I've tried this for EM on several versions, and they tend to stick to whatever they're given first. That goes down, they stop working even though other IPs they are given in the RR list are still alive and happy.

2) Both web services (i.e. both UCCX servers) responds perfectly (from a network point of view - they return data) so the phones will see both of them as 'up'.

So you really need some intelligence in the path to determine which is the correct service, and if you don't have a device that can do it, the intelligence needs to be between the agent's ears... or at least their Supervisor's.

I do agree that it could be designed much better... if the phones worked with RR DNS, EM fault tolerance would be easier to set up, and that combined with an option to have the inactive IPPA service actually inactive (i.e. not listening on the TCP port) would give a slicker solution.

Aaron

Please rate helpful posts...

Actions

This Discussion