Faster failover for K8s on containers #1931
Description
Is your feature request related to a problem? Please describe.
When nodes share the same kernel, all nodes agent-ha try to failover and contend until the correct one wins.
Describe the solution you'd like
When we detect the path going down we could check if the hostnqn is for the pod node, and if not simply ignore it.
Describe alternatives you've considered
The cluster agent could reject the incorrect nodes quicker, but this still leaves ha-node confused
Additional context
Easy to repro, just try to test ha in a kind cluster...