cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
4449
Views
5
Helpful
12
Replies

CUCM publisher subscriber call processing failover

J_Vansen_S
Level 3
Level 3

Hi All,

I need some opinions on the failover roll of pub & sub, and how best to design failover based on their setup.

We have 2 units of BE6k:-

  • one in HQ, and
  • another in RemoteSiteA

Total of 900 Ip phones:-

  • 700qty in HQ,
  • 200qty in RemoteSiteA

Each site has a voice gateway for E1/T1 telco link, where telco will forward calls in parallel

We will be doing cluster over WAN, for the CUCM, CUCIMP, CUC, UCCX.

Since we only have 2 boxes,we are looking installing Publisher at HQ and Subscriber at RemoteSiteA

The question is; which should be the ideal unit to take on the role for call processing & device registration?

or should both Pub & Sub do the load sharing?

or should HQ ip phones registers to Pub and Remotesite ipphones registers to Sub~?

In terms of maintenance wise, would load sharing pub+sub = hassle+tedious

What is the best practise in my scenario?

Appreciate any insight

2 Accepted Solutions

Accepted Solutions

Chris Deren
Hall of Fame
Hall of Fame
Configure 2 CUCM groups with opposite CM server order, assign these accordingly to site specific device pool. Basically you will load balance traffic between the 2 CUCM server, where HQ users will use HQ Pub as primary, and remote users their local server locally. Alternatively if you have room on your UCS server you can deploy another CUCM Sub at the HQ location and use that as primary call processor for HQ users. You can also apply the same logic with UCXN cluster by defining different voicemail profile/pilots/RPs and point to local Unity Connection server which would eliminate the need for sending voicemail traffic across WAN.

View solution in original post

No as the CUCM nodes are not licensed in releases of CUCM and you can have up to 8 call processing servers in a cluster.

View solution in original post

12 Replies 12

Manish Gogna
Cisco Employee
Cisco Employee

Hi,

In ideal scenario you don't have Publisher doing the call processing and phones are registered with subscribers for call processing. However, in your case since there is only one server at each site you can register the phones to local servers at each location and have the server of the other location as a backup for TFTP in the option 150 and call processing in the callmanager group. Since your setup is already close to the limit of max devices in BE6K you don't have to worry about any scalability or performance issues once this setup starts functioning normally in this way.

Manish

- Do rate helpful posts -

Thanks Manish for the info.

If i understands your info correctly, you are sugguesting for PUB to do the call processing + TFTP for HQ, and SUB to do the call processing+TFTP for RemoteSite. That both backups each other as secondary?

That is correct, same as pointed by Chris as well [+5]

Manish

Awesome! Thanks

On a side note, is it possible for the servers to swap role, eg Publisher to Sub, and Sub to Pub when its already fully installed with phones registered?

Reason being, at the moment HQ site is not physically ready, ie server room.

But RemoteSite is ready and that i can start server installation there. Which means ill need to install Pub at remoteSite for the start, then work on HQ installation when the site is ready.


Appreciate your advise

Unfortunately there is no way to do that, your best bet might be to re-host the pub and move the virtual machine at that point.

hi Manish,

So ideally i would turn on all the services available to my publisher, and do the same at on my subscriber correct?

We're looking at services like CallManager, CTIManager, TFTP, DirSync, AXL, TAPS turned on, on both pub and sub?

Hi,

In your scenario with two servers in the cluster and one server at each location this would be the correct option.

Manish

Do not turn on all services, only start services you need so that you do not waste CPU cycles on things you do not need.

Chris Deren
Hall of Fame
Hall of Fame
Configure 2 CUCM groups with opposite CM server order, assign these accordingly to site specific device pool. Basically you will load balance traffic between the 2 CUCM server, where HQ users will use HQ Pub as primary, and remote users their local server locally. Alternatively if you have room on your UCS server you can deploy another CUCM Sub at the HQ location and use that as primary call processor for HQ users. You can also apply the same logic with UCXN cluster by defining different voicemail profile/pilots/RPs and point to local Unity Connection server which would eliminate the need for sending voicemail traffic across WAN.

hi Chris,

Thanks for your valuable input. Its a great idea to have a 2nd sub in HQ. However we have bought only 2 sets of CUCM licenses/cuwl. Does that limit us in having a 2nd sub in HQ?

No as the CUCM nodes are not licensed in releases of CUCM and you can have up to 8 call processing servers in a cluster.

Hi,

Do remember that you can add additional BE6000 servers in a cluster, however a deployment of BE 6000 solution on UCS C200 Rack-Mount Servers in a cluster exceeding 1,000 users and 1,200 configured devices is considered a regular Unified CM cluster, and as such it is bound by all pricing and licensing requirements and design guidance for regular Unified CM clusters and UC on UCS deployments.

Manish