Richard Bucker

Bare Metal Warning

Posted at — Sep 26, 2022

HA HA HA. I’ve been talking about Bare Metal vs containers and virtualization for a while… again. I still do not have much good to say about virtualization except that orchestration can be fun and profitable. But here’s the thing.

Bare metal can fail to reboot

I just performed a syspatch on a bare metal OpenBSD installed MiniForum machine and in the throws of the patch I realized that this was the second time I was patching the same packages. I cannot say for sure what happened or why but it seemed like twice. That followed up with a ssh reboot command… and for some reason the machine never came back. And so all that DR math rushed back into mind.

The more machines in a network the more likely something is going to fail.

The more things fail the more likely that reparations are going to fail too… Take that blue/green servers during an upgrade. The blue server fails and green takes over. Now you have a different competition for survival. Get some hands and eyes to restart blue or start restoring green on a third machine.

three is the magic number

You need all three machines in order to give all parties a chance to experience all the failures. Sometimes as a simulation and others as real events. Practice makes perfect.