HA deployment: use Rancher to manage Rancher

Rancher 1.x
1

After looking at the Rancher HA set up, I wondered why I couldn’t just use Rancher. So basically, I’d have the managing Rancher instance running with just one stack that had one Rancher instance pointing to a host running mysql. And I’d have an extra host on standby. I know this is not truely redundant, but the Boss Rancher would restart the worker Rancher if it died or health checks failed, and if I had a hardware failure, Rancher would just start the worker Rancher on the standby host.

I really don’t want to mess with redis, zookeeper, etc. This is definitely the lazy man’s HA, but is there any reason this wouldn’t work?

2

Rancher-on-rancher (or -in-) is definitely a way we could do automatically managed HA eventually.

You would need an external (i.e. not the one in rancher/server) MySQL container or hosted-as-a-service (like Amazon RDS) since that stores all the state and will go with the failed host. Otherwise this sounds like it would work, though of course you just move the problem one turtle down and the “boss” one has no redundancy…