Achieving HA Kubernetes Cluster with Combined Control Plane and Worker Nodes

br flag

I have a question regarding the possibility of achieving a high-availability (HA) Kubernetes cluster by combining three nodes that serve both as control plane and worker nodes. My understanding is that each worker node can connect to its respective control plane.

I would like to know if any additional setup is required for this configuration to work effectively. The reason I am exploring this option is that I am hosting my cluster on cheap VPS instances where I do not have access to a load balancer. Therefore, I am looking for an alternative approach to achieve HA without relying on a load balancer.

The cluster does not need to be real HA, but it should be possible to reboot single nodes for maintenance work without failing the whole cluster.

halfer avatar
cn flag
Not an expert here, but I think this is what MicroK8S does by default. I've installed Ubuntu Server + MicroK8S on three nodes, it snapped into HA mode automatically, and each node hosts a worker and a control plane. There is a special mode one can select if one wants a node to be worker only, but it's not the standard setting.
Fritz avatar
br flag
@halfer thanks, i'll have a look. I'm not sure how or what is automatically happening from my research. I'm using kubeadm so everything is different again
I sit in a Tesla and translated this thread with Ai:


Post an answer

Most people don’t grasp that asking a lot of questions unlocks learning and improves interpersonal bonding. In Alison’s studies, for example, though people could accurately recall how many questions had been asked in their conversations, they didn’t intuit the link between questions and liking. Across four studies, in which participants were engaged in conversations themselves or read transcripts of others’ conversations, people tended not to realize that question asking would influence—or had influenced—the level of amity between the conversationalists.