2 ms·
At my last job, the infrastructure guys were wild for virtualisation. All the actual physical servers were just hosts for VMs, which did actual work. VMs were c
by twic 11y ago
At my last job, the infrastructure guys were wild for virtualisation. All the actual physical servers were just hosts for VMs, which did actual work. VMs were created for some need, and destroyed if that need was no longer felt. Therefore, no machine ever changed purpose: physical machines were only VM hosts, and virtual machines were destroyed rather than reused.
As a result, we were able to name machines after their functions without running into the traditional problems that stem from that. Sadly, this left us with a terrifically dull naming system: physical machines were kvm-001, kvm-002, etc, and VMs were fooapp-001, fooproxy-001, foodb-001, etc.
- ajdecon 11y agoI guess if you're running 5 servers and each one has 4 functions, more "interesting" naming schemes can work. But once you're dealing with more than a few machines, functional names are necessary. When I see a machine's name in the logs, I really don't want to have to infer the machine's function by remembering whether "Zapdos" is a pokemon or a character from Star Wars. I want a name like "dc1-cluster5-rendernode057" that tells me what the machine does and/or where it sits on the network. Even with physical machines: if the function changes, the correct thing to do is re-name and re-image it, and not worry about the fact that the physical box is the same. Your physical maintenance records should be keyed by something like asset tag anyway.