Our data center move checklist had 3 major parts - what to do before the move, what to do a short time before and at the time of the move, and how to validate that everything was up and running correctly. The advance portion of it included things like notifying customers of the planned downtime, figuring out who’s DNS was where so it could be properly modified at the time of the move, and quite a few others.

One item caught in advance using this checklist, for example, was that our own network-accessible power strips (for remote rebooting, etc) had the wrong kind of plug to interface with the data center’s power distribution setup - we needed to get adapters. This is an example of one of those things you definitely want to know in advance so you’re not standing in the data center late at night with no way to power your gear.

The “during” portion had a complete checklist for both the tear down at the old location and the build out at the new one. It was ordered sequentially, with the idea of maximum speed in mind while preserving systems integrity. The tear down portion involved readdressing and clean shutdowns of each machine in the correct order, followed by removing equipment from the racks and shuttling them out to waiting transportation, followed by cleanup of cables, etc.

The build out portion had the order listed out to bring the systems back online as quickly as possible, and designated specific portions of it to individuals, ideally timed so that as machines were mounted and cabled they could quickly be turned back on and checked.

Once the build out was finished the final part of the checklist was run through, and we made sure each service was back online. Once finished, it was time to clean up and go home. There was some more follow up work in that part of the checklist that was planned for the next day, and was taken care of then. The first move had a team of 2; the second move expanded up to a team of 4. A good rule of thumb in this situation seemed to be correct - more than two people per rack is probably too many. The second move, with twice as much equipment took about twice as long as the first, so in our case it scaled pretty linearly.

The data center we chose has been performing spectacularly since our last move. We haven’t had any downtime, support has been excellent, and the security arrangements and access policies allow us to get in to work whenever we wish. As we’ve added new racks they quickly take care of their end and make sure they can meet our future needs. The moves may have been a little painful, but we’re happy knowing that we shouldn’t have to do it again.

The decision to move your IT infrastructure from one data center to another is not one that ever should be taken lightly. There can be a considerable investment in planning time and personnel, and there is usually some downtime unless you have purchased twice as much equipment as is needed to do a live cutover. However, by planning it well you can minimize the negative factors and make the transition painless.

We ended up needing to move data centers twice in one year, and our planning made both moves go as smoothly as possible. Our first move was made necessary because of the data center’s policies. Some of these policies (which I won’t get into here) made it difficult or impossible to get our work done. After researching other nearby data centers we moved to a new one, but within a few months it was obvious this data center wasn’t meeting our needs. Although we didn’t want to move again we didn’t really have a choice.

At the time of our first move we had 1 rack of equipment, which had both a mix of our servers and network gear and customer equipment. The services included email, web servers, terminal services, and others, and needed to be up as close to 100% as possible. By the time we moved again we had expanded to two racks, and had started allotting serious downtime from periodic network failures that the data center support personnel could not identify or clear up. As time went on the situation became intolerable and required the second move. In both moves we planned well in advance.

The key item during planning was building a comprehensive checklist. This checklist included every possible aspect of the move that I could think of, from planning our address allocations and usage, notifying customers, dealing with DNS, and quite a few others. I’ve found that in dealing with large, complex tasks (such as moving to a new data center) a growing checklist, planned out in order, helps keep you on track, makes sure you don’t miss anything, and, when you’re moving servers and readdressing at 3:00 AM and incredibly tired, keeps you from losing track of what you have and haven’t already done and helps you avoid making serious mistakes. Next time I’ll get into what we put into the checklist.