Naming your servers, Anchor-style

By April 16, 2012Technical

If you’ve been an Anchor customer for a while, you’ve probably heard the names of a few of our servers. Since the company’s inception way back when, we’ve been using fun nautical-themed names for everything. That’s why our backup servers, being the disk-hogging monsters that they are, have names of great sea beasts like leviathan and cthulhu.

This isn’t just us being silly, there’s sound logic behind it. Choosing good names for computers is covered by RFC 1178, it offers advice on things you should and shouldn’t do, and explains why. Even having been written over 20 years ago, the guidelines are as relevant as ever. It largely revolves around being clear, unambiguous, easy to remember and reducing the chance of mistakes in usage.

This has served us very well for over a decade, but we’ve recently made a decision to evolve our policy a little further to meet our needs.

One of the things you learn as a hosting company is that hardware sucks. Some hardware sucks less than other hardware, but it’s pretty much universal. Eventually you’ll have a piece of hardware give up the ghost and you need to replace it. This is a well-practised procedure for us, and it usually involves moving all the disks into a standby chassis (you do keep spares, don’t you?).

Now you have an running system, perhaps named “whiskey“, inside a chassis named “rum“, which is a bit of a problem. You can’t go and rename the operating system, so our procedure so far been to rename the hardware assets. However, this involves a lot of changes in multiple systems, and it’s easy to make a mistake and get things into an inconsistent state.

We’ve decided to solve this by having separate names for hardware (the chassis) and software (the operating system and associated DNS records, etc). There’s a clear logical division between these realms, which cleanly aligns with the way we deal with hardware and software.

The primary benefit is that a name is assigned once and Never Ever Changes. Think about the the problems this solves:

  • When you replace a failed chassis with a spare, you re-map the OS to the new hardware, and you’re done
  • A hardware-upgrade, likewise, requires minimal updates to your records
  • It’s easy to refer to a piece of hardware before you even install the OS on it
  • Giving a previously-used chassis to a new customer is less confusing. No more, “Cocos? You mean back when Alpha Corp was using it, or after Beta Group got their hands on it?

In keeping with the nautical theming, we’ve chosen to assign names to hardware based on the chassis-type. Each model of Dell server (R410, R510, R710 etc.) gets a name, and each individual piece of hardware in the series gets a numeric suffix. Our remaining Supermicro boxes will constitute a single series, as it’s not worth bothering with different names for the many different models.

Our DC technicians look forward to having all the records and labelling updated in the near future, then they can breathe a sigh of relief and restore some of the sanity lost over the years. If you’ve got any fun server-naming stories we’d like to hear them. Have you run into the same problems, and perhaps even solved them before?