Friday was Canada Day, Monday is Independence Day. That makes it a very long weekend when you take stat holidays in Canada and report into a group in the US ($DAYJOB). As a result, I get to spend the Monday collecting any items that cropped up on the Friday and taking care of them before my compatriots south of the border have a look at them. This doesn’t really happen all that often, but it’s nice when it does.

Interestingly enough, my contract ends on July 5th, so I guess I’ll hear about an extension or not pretty much at the last possible moment. Either way, I’ll have things to do for six more months or I’ll have some time off while I decide what to do next. I do have a number of things I’d like to roll out at $DAYJOB to make life easier, but if I’m not going to be there, I won’t be doing it. I’ve been kind of dragging my feet over rolling out a configuration management (CM) tool, as the existing systems are all running the approved software, but there will be an upgrade at some unknown point in the future which would be the correct time to introduce the concept of a CM system.

The biggest problem is that OS upgrades and configuration decisions only happen when it’s time to do a change to the existing software, which can be a year or two in the making. Not having a surplus of systems to test on also makes it challenging. I have managed to keep a couple of systems that were to be retired as test units this time around and things look promising. If I get the extension, I’ll probably test both CFengine and Puppet as see which will be easier to turn over to the rest of the team for future upgrades. The nice thing will be that any machine tagged as “TypeX” or “Type Y” and managed under a CM system should always conform to the accepted standard, making rollout easier. I did look at this in the past, but the configurations were so generic that a standard install with a couple of services disabled was sufficient and the investment in a CM was not warranted. The newer environment makes distinctions on machine purpose so a CM appears to have worth. Hopefully I’ll be able to test out some ideas and do a comparison. I think that will be the right thing to do. Either way, the next rollout will be staggered across the machines as opportunity permits, which removes any real incentive to create such a system.

Of course, being the kind of systems administrator that I am, any opportunity to automate should be examined. This will pay off, but it won’t have the impact as a full upgrade would, so justifying the time to do it is a little harder. At this point, it only helps me if I have to do multiple units. The payoff is when I have to roll out a number of systems. Hopefully before I get this running, as that would make the rollout smooth, fast and efficient, as well as cover off the systems that have unique configurations.

I guess this post really counts as thinking out loud.