Upgrading a critical enterprise call processing system to a completely new virtualized server platform sounds pretty tricky. Doing it from 5,000 miles away, in the public square of a sleepy Spanish village using your laptop and a VPN connection over the free municipal WiFi service sounds … well, maybe a little crazy. Recently, I did just that, migrating our Cisco Unified Communications Manager (UCM) cluster in Johannesburg, South Africa to the Cisco Unified Computing System (UCS) platform from the small village in Spain where I was vacationing.
Normally I don’t work while on vacation, but it’s hard to schedule an approved time for critical system upgrades, and once you have it, you don’t let it go. Scheduling time for personal vacation isn’t easy either so both ended up on my agenda. Here’s my view of the village square while I was doing the upgrade – still too early even to get myself a café con leche.
While sitting on the terraza, I used the Cisco Any Connect VPN client on my laptop to establish a secure connection to the Cisco network and from there to the systems in Johannesburg. Working with Jason Daniels, a colleague in Cisco's Barcelona office, we performed remotely all of the tasks for migrating the software and configurations for both the Cisco UCM cluster and a Cisco Unified Presence Server pair to the new Cisco Unified Computing System (UCS) hardware at the Johannesburg site. (The physical servers and network connection had been installed and tested several weeks before our scheduled migration date.)
These tasks included:
- Copying the new software images to local storage on the UCS servers.
- Installing and restoring the Cisco UCM Publisher on the UCS servers, after shutting down the Publisher on the Cisco Media Convergence Servers that were currently in service. This step was necessary because we used the same hostname and IP address for the new servers. Call processing by the Cisco UCM Subscribers continued without interruption during this step.
- Building and restoring the new virtualized subscriber nodes on the UCS servers, then transferring traffic and shutting down the old servers.
The process was quite smooth and in some ways we forgot that we were working in a virtualized environment. But the move to virtual servers has meant some changes for ongoing operational procedures. Cisco IT operations engineers who previously had responsibility for server hardware now must work with separate teams that manage the server hardware and the virtualization software. We've needed to clarify our procedures for identifying the real source of the problem and who should get called to fix it. We have also trained our engineers about how some other routine tasks, such as mounting disk images, have new procedures in a virtualized environment. The Johannesburg cluster is the first full CUCM cluster migration in Cisco IT’s plans to move all of the company’s Cisco UCM clusters to the Cisco UCS platform. For the remaining clusters, although Cisco IT will perform many migration tasks remotely, it's likely that the IT engineers will do this work from Cisco offices, not a small-town café. But based on my experience, they could if they wanted to.