K8s High Availability HaProxy setup

#1

Cluster information:

Kubernetes version: V1.14
Cloud being used: bare-metal
Installation method: APT package manager
Host OS: 18.04

In a Kubernetes bare-metal HA setup where an HaProxy lives between the master and worker nodes, what are the consequences if the HaPRoxy goes down and communication is lost between the worker and master nodes?

Looking @ the stacked etc topology: Options for Highly Available Topology - Kubernetes

0 Likes

#2

It will most likely result in a similar behaviour as if your master nodes went down.
https://kubernetes.io/docs/tasks/debug-application-cluster/debug-cluster/#specific-scenarios

Whatever you use as an LB in that HA setup should itself be HA, otherwise you have just moved the single point of failure away from Kubernetes itself to something else.

0 Likes