How do we cleanup the static ip incase of blade H/W failure of worker node. about danm HOT 3 CLOSED

that has nothing to do with DANM. Kubernetes works in a central server - remote agents setup, Kubelet being the node agent.
Kubelet is the one invoking the CNI interface.

If you stop Kubelet, there is no one to invoke the CNI plugin, therefore no network provisioning operations can happen on the node

from danm.

@Levovar Yes I can understand, But there will be a downtime until that machine comes up.
Just a suggestion Can we write some webhook in admission controller to check the Danm endpoints and check the pods running status take the decision to cleanup the ipam.

Any suggestion ?

from danm.

there will be a downtime, but it does not necessarily mean that the Pod having the static IP is also down.
even if the Docker daemon is not running, the container still can.
I'm not in favour or reaping these IPs, because if the Pod having it assigned is still running, and we re-assign the same IP to another Pod at the same time, that will lead to problems.

On the flip side, when Kubelet comes back online, it will synch the status of the Pods. If a Pod died while it was down, I think / expect Kubelet to invoke a CNI_DEL for the Pod.
In that case the static IP will be freed.

from danm.