[feasibility-research] Handle machine failure

[gaocegege]

According to the docs about restart policy here: https://kubernetes.io/docs/concepts/workloads/pods/pod-lifecycle/#restart-policy, once bound to a node, a Pod will never be rebound to another node.

Thus when the machine node is down, we should try to handle such a failure.

There are some options to achieve it:

• Use descheduler
• Implement the logic in the operator.

I will try to investigate the cost of this feature.

[jlewi]

@gaocegege Doesn't the CR already handle this? Won't the reconcile logic detect that the pod isn't running and create a new one.

@johnugeorge @richardsliu

[gaocegege]

I think it cannot solve the Split brain problem.

[jtfogarty]

/area engprod
/priority p2

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.