0

when running pod-workload on an AKS-cluster with autoscaling enabled, what happens, if the pool gets scaled down? If a Pod is running workload on a node, and this node is being removed because of nodepool-scaledown, how does Kubernetes manage to keep my pod running? Even if there is a re-scheduling, this would interrupt my workload, which would lead to a failure in my case.

So how does K8s know, when a Node is "ready" to be scaled down ?

Improve this question

1 Answer 1

Reset to default
0

If a node is being gracefully removed then Kubernetes will inform pods that they are about to be killed and provide a window (30s I think) to gracefully terminate before it is killed. During this window your service can create another pod on a different node to take over from the one being killed.

Improve this answer
3
  • Hi, thanks for your response! So how ist this "transfer" made? Doesn't the application that's running on the pod/container, need to have a logic to get through this process of transferring from container-engine on node1 to container-engine on node2 ?
    – greterpiffin
    Oct 2 at 17:40
  • As far as Kubernetes is concerned there's no transfer, the pod is killed on one node, and restarted on another. So if there is some sort of logic required to pickup a job where it left off on the other node then this has to be handled by the applicaiton.
    – Sam Cogan
    Oct 3 at 8:27
  • ok, great. Thanks!
    – greterpiffin
    Oct 3 at 10:02

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .