5 server UCM cluster - how to recover from Publisher failure.

[sdoherty]

We have a 5 server cluster - in a scenario where the publisher has a failure will one of the subscribers assume the role of publisher until the pub can be restored?  What is the procedure?  Version is 7.1.5

Thank you

[Rob Huffman]

Hi there,

You cannot promote a Sub to Pub in CUCM so this is where

the DRS (Disaster recovery System) comes into play. Here's a doc

with some pretty good step by step directions;

http://www.cisco.com/en/US/docs/voice_ip_comm/cucm/drs/7_1_2/drsag712.html#wp72076

Cheers!

Rob

PS: make sure to have all your CUCM media including SR's and custom files on hand

just in case.

[sdoherty]

Thanks Rob.  According to the people in our VOIP lab they have two server - one a pub one a sub.  They turned off the pub and the sub took over.  They checked the cluster and it showed as the pub.  So this seems to be the way it worked.  Then they made a change on the new pub and brought up the old one and the change transfered and then the pub that was powered down took over the pub role.

Is this the way it works?

[Rob Huffman]

Hi there,

Well...they are correct in a sense, in that the phones and call processing don't

cease just because the Publisher has gone down. But the Sub can never become the

Pub without a re-build

The Subscriber only has "write" access to these database changes;

The configuration database is stored on a publisher server, and a read-only copy is replicated to the subscriber members of the cluster. Most of the database changes are made on the publisher and are then communicated to the subscriber databases, thus ensuring that the configuration is consistent across the members of the cluster and facilitating spatial redundancy of the database.

#

Database modifications for user-facing call processing features are made on the subscriber servers to which the IP phones are registered. The subscriber servers then replicate these database modifications to all the other servers in the cluster, thus providing redundancy for the user-facing features. These features include:

#

•Call Forward All (CFA)

#

•Message waiting indicator (MWI)

#

•Privacy Enable/Disable

#

•Extension Mobility login/logout

#

•Hunt Group login/logout

#

•Device Mobility

#

•Certificate Authority Proxy Function (CAPF) status for end users and applications users

#

•Credential hacking and authentication

http://www.cisco.com/en/US/docs/voice_ip_comm/cucm/srnd/7x/callpros.html

Cheers!

Rob

[sdoherty]

OK, thanks Rob.

SO if I have our main building and it goes offline with the publisher for say 2 weeks due to flooding or whatever and at our backup site we have 2 of the 5 servers in the cluster - both subscribers, can we wait for the publisher to come backup or should we rebuild it at the backup site?

We have a spare server at our backup site.

Thank You

[Rob Huffman]

Hi there,

That is an excellent question

I've never seen a hard and fast rule for CUCM and the Pub being offline for

that period of time. I would think you'd probably be OK if you can

live with no dB changes that require the Pub. You'd have to do a cluster

re-boot for sure when it came back online, but this is just my opinion.

We did have a Unity Connection Pub down for about 10 days when working with

TAC on a crazy bug, but that may be comparing Apples to Zebras

You might want to check with TAC to see what they have to say.

Cheers!

Rob