I'm putting together a bare metal Kubernetes cluster using RKE2 and Rancher, and I'm a bit confused about the best configuration for my master nodes. I can either go with three master nodes or a setup with one master and two workers. My main concern is maintaining quorum during maintenance or unexpected issues like switch upgrades or power failures. If I choose the 3-master configuration, will I risk losing quorum if I reboot the switch? On the other hand, a single master would mean I'd sacrifice high availability, but at least I wouldn't have to worry about quorum loss during those times. Would my workloads still run normally if I had to reboot the master node? Sorry if this question seems basic, but I couldn't find relevant information.
1 Answer
If your control plane goes down, your running workloads will keep operating, but you won't be able to schedule new pods until it's back online. So, a 3-master setup is better for availability. The downside is that you’ll need more resources for those additional control plane nodes. It's worth considering, especially for redundancy.
Thanks for the clarification! I appreciate it. Also, I'm curious about what happens if all three etcd nodes lose connection. Will they all go read-only? Is there a risk of data corruption in that scenario?