Prime Collaboration Deployment (PCD) for CUCM 7.1.5 migration

[syed shahid]

Hello Folks,

We will be migrating our CUCM 7.1.5/MCS platform over to CUCM 10.5.x/UCS.  I have reviewed the PCD Administration Guide Rel 10.0.1.  We will be using the simple migration and maintain same IP scheme. Our cluster is split across WAN - located at 2 datacenters.

I wanted to take following approach because this way we can prepare the new platform outside of maintenance window and cut the down time drastically.

1) Implement the PCD server on the production network - WAN connection.

2) Run the discovery task and complete the process to discover all the servers in the cluster.

3) Disconnect the PCD from WAN.

4) On an island network, connect PCD and the new UCS/VMs.

5) Prepare the destination cluster via the Migration Task.

6) Because the existing cluster is at 2 sites, we may ship the UCS hardware directly at those sites accordingly.  For that purpose, I am thinking to run PCD application from each site separately and basically have two identical copies of discovered cluster.

6a) From each site I will then run the steps 4/5 above, and only assign specific servers in the migration task, those to be located at that site which can be done by selecting specific servers. OR

6b) Alternatively, I may run one PCD instance only, run the migration on specific servers at the site and then take a backup of the PCD server. Load this backup image at the second site and then build the remaining servers for that site only (island network).

7) In the new environment, stop the callmanager service on SUBs and TFTP service on the TFTP server.

8) During the maintenance window, manually drop network connections (port shutdown etc.) for the production servers

9) Bring the new UCS servers live over WAN and wait for the cluster communication/db replications completed.

10) Start the callmanager and TFTP services.

I have am hoping to get some help on confirmation for the following:

1) Would PCD be able to run the migration task while not over the WAN, i.e. it does not have a live connection with production cluster. I assume this should be possible because it has already completed the discovery.

2) Option 6 or 6a, either of them, or none of them will work? In that case we should ship the UCS at one site and build the cluster before shipping out to each site.

3) Any other comments.

Thanks a lot!

[Jaime Valencia]

Discovery means just that, it will find out how many servers you have in your cluster and list them, nothing else. You NEED to have a connection with the old cluster for the migration procedure so that it can pull all the information from them. What you're proposing simply won't work. That being said, you probably want to go back to the drawing board.

As your step #6 was based on the premise you thought it was just a matter of discovering the cluster to make those happen, they obviously won't work.

If you want to do a similar procedure to what you mention, you probably want to use the jump upgrade, go to 9.1(2), then upgrade to 10, or re-do your whole PCD plan.

[syed shahid]

Well, it does more than finding and listing the servers as noted in the  Administration Guide:  "When the Discover Cluster operation is performed, the Cisco Prime Collaboration Deployment server contacts the publisher of the cluster and retrieves the cluster information. It then contacts each server individually, installs a cop file on the server to retrieve configuration information, and collects information about the hostname, IP, product type, and both active and inactive versions for that server."

But I get your point that during discovery it does not take the dr backup of the cluster. So the provisioning database is not copied to PCD server. That happens during migration task directly between source and destination servers. If so, would it be possible to build the servers using migration task, and then do the DR Restore as another step?  Basically build the servers with PCD tool and then do the restore manually?

Just your thoughts although I doubt this would be possible because the PCD may not mark the process as complete thus the destination cluster may not be useable or ready for DR Restore.

[Aman Soi]

Hi Syed,

[+5] to Jaime.

As far as I know, once u discover the existing cluster, a platform cop file is created which is used later for migration.

During the migration Task of Destination Cluster, you do not get any such option of DRS Restore .The only options which u get are for change of IP address/Hostname.

You do not get any such option to do this.

regds,

aman

[syed shahid]

Thanks for response. Much appreciated.

[Jaime Valencia]

No, that's not possible, the actual term for this isn't even DRS, the "backup" PCD takes from the old cluster is only hosted within PCD itself during the time it installs the new cluster and when it pushes the info to it, and it's not even user accessible.