CUCM: Redundancy (Call processing failover)

The result of the Cisco IP phones failover to secondary or tertiary CUCM server is instantaneous in the point of view of enduser, if everything was configured correctly. The test was done with 13 registered devices, of course as the number of registered devices increase the recovery time may vary. Only SCCP and MGCP (used for Cisco 2951 router which is a voice gateway with 4x FXO) are used, no SIP.

A cluster also known as Cisco unified CM group contains one or more CUCM, each cluster supports up to three CUCM servers. Every cluster has only one publisher and the rest is subscribers. You only need to do configuration in publisher, however you need to turn on Cisco CallManager service on every subscribers not only on publisher.

A publisher is also known as the first node, a subscriber is also known as subsequent node. Only the publisher needs to synchronize its clock with the lowest stratum NTP server, subscribers will get their time through NTP from publisher only.

A device pool is associated to a Cisco unified CM group which in turn also attached to the phones. By default Cisco unified CM disable auto-registration, however for huge deployment of many IP phones you can enable auto-registration first then disable it after all IP phones are registered.

IP phones also known as registered devices will register to the primary CUCM server based on what was configured in Cisco unified CM group. In other words, all IP phones which have the same device pool will register themselves to the same CUCM server.

In the Cisco Unified CM group, the top of the list is the primary, the subsequent item of the list is the secondary, the lowest item in the list is the tertiary.

It is possible to do call processing distribution by creating two or more Cisco unified CM groups and two device pools. In this setup, I have two CUCM servers, two device pools are created and two Cisco unified CM groups are created, each Cisco Unified CM groups have reverse orders on which is the primary CUCM server.

Turn on Cisco CallManager service on subscriber
All services are disabled by default on subscriber, services turned on on publisher is not replicated to the subscriber in order to create cluster you need to turn on Cisco CallManager service on subscriber.

To turn on the service, on subscriber go to Cisco Unified Serviceability, click on Tool > Service Activation.

Click on Tools.
Select the services that are also used by the publisher.

Cisco Unified CM group

Create two Cisco unified CM groups, from CUCM, System > Cisco Unified CM group.

CM_CUCM is the primary, CM_CUCM1 is the secondary. The top item of the list is always primary.
The next group has a reverse order than the first. CM_CUCM1 is the primary, CM_CUCM is the secondary. The top item of the list is always primary.

The Cisco unified CM groups are created automatically when you have provisioned the servers in System > Server from CUCM. However you need to manually add the CUCM servers in order yourself.

The publisher server is created during the installation, you need to add subscriber before installing the subscriber.

Publisher and subscriber servers.
Server configuration.

CUCM is complicated because it has too many dependencies, creating a cluster is one fine example. In order to create Cisco unified CM group, it depends on Cisco Unified CM, in order to create Cisco Unified CM it depends on Server.

Note: Many IP phones failed to failover to the secondary CUCM server is likely due to DNS issue, the DNS has no record of the CUCM hostname mapped to an IP address.

Another issue I encountered is to use hostname in System > Server despite I have no DNS. If you have no DNS servers please put in the IPv4 address of the servers instead. If you insist on using hostname despite you do not have hostname you will find the failover to be so slow that it is unacceptable for factory acceptance test, your IP phones will not register to the CUCM server as expected, in conclusion the result is erratic and the failover is slow and sometime IP phones are not registering to secondary CUCM server at all!

Device Pool
My setup is to create two device pools and two Cisco unified CM groups, so that some IP phones join the first cluster and some phones join the second cluster to form call processing distribution.

Two customized device pools are created.
DP-1 device pool, I have included media resource group list into the device pool. In the media resource group I have added the two Cisco unified cm groups.DP-1 is associated with FAT1 cisco unified cm group.
The second part of DP-1 device pool.
DP-2 device group is associated with FAT2 cisco unified cm group, media resource group list attached.
second part of DP-2.

DHCP servers and DHCP subnets
I have heard that only the publisher has to be the TFTP and dhcp server. In this setup I have included two dhcp servers which are publisher and subscriber which I am not sure if this is correct or not.

Two DHCP servers are created.
DHCP server internal configuration. Basically I have only defined the tftp server ip address, and this DHCP server profile is used to assign the IP address pool.
Two DHCP subnets are created.
DHCP subnet internal configuration.

Service parameter adjustment

From CUCM, System > Service Parameters

Choose the publisher and choose Cisco CallManager service, the parameters will appear.
Station and backup server keepalive interval was at 60seconds by default, I changed to the minimum – 10s. Station keepalive interval was 30seconds by default, I changed to 10s which is the minimum.

Station and backup server are referring to primary and secondary CUCM servers.

Apply device pool to the IP phones

From CUCM, Device > Phone, you can attach individual IP phones to device pools and they will register to the primary CUCM server as specified in the Cisco Unified CM group arrangement.

There are Cisco VG224 and VG204 as well, the primary and secondary CUCM server are not determined by device pool instead the primary and secondary CUCM servers are determined by Cisco VG command line. The IP phones are however influenced by the device pool.
When the CUCM publisher is switched off, the IP phones that were registered in publisher switch over to the subscriber, as shown in the picture all phones are registered in subscriber.
When the CUCM publisher is booted up and ready, all IP phones which were previously registered in publisher switch back to publisher again.

IP phones that are registered in publisher will switch back to publisher from subscriber once publisher is up again, similarly IP phones that are registered in subscriber will switch back to subscriber from the publisher once the subscriber is up again. The failover is seamless in the point of view of user.

Cluster verification with Cisco unified reporting

You can generate report in Cisco Unified reporting to check if the CUCM servers have discrepancy within the same cluster.


6 thoughts on “CUCM: Redundancy (Call processing failover)

  1. Nice tutorial. I’m finding this a bit late; are you planning on any updates that coincide with the new “collaborations” track?

  2. CUCM group and cluster are not synonymous. Cluster can have up to 9 servers. Group can have up to 3.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s