Describe the problem/challenge you have
Current polling interval in NodeController is hardcoded to be 60 seconds. This means that the latency for node updates, and in turn CSIStorageCapacity updates, is up to 60s, even after a PVC was bound. As a result, custom k8s schedulers that are aware of storage can make bad decisions - either unnecessary evictions, or impossible allocations.
Describe the solution you'd like
Make pollInterval in lvmnode.NodeController configurable, with the default being 60s like today, and allow users to customize according to their use-case.
Describe the problem/challenge you have Current polling interval in NodeController is hardcoded to be 60 seconds. This means that the latency for node updates, and in turn CSIStorageCapacity updates, is up to 60s, even after a PVC was bound. As a result, custom k8s schedulers that are aware of storage can make bad decisions - either unnecessary evictions, or impossible allocations.
Describe the solution you'd like Make pollInterval in
lvmnode.NodeController
configurable, with the default being 60s like today, and allow users to customize according to their use-case.Environment:
kubectl version
): v1.27.4/etc/os-release
):