You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
We could possibly improve the behavior of nfd-gc by removing stale NRT objects (whose creator pod does not exist anymore). The ownerReference field shouldn't be used as that creates warnings (see k8s garbage-collection). We could use an nfd-specific annotation (like nfd.node.kubernetes.io/owner-pod: <namespace>/<pod-uuid>). The nfd-gc would delete the object if the pod doesn't exist. Possibly acting only if <namespace> is the same where nfd-gc itself is running (so that it doesn't need cluster-wide access to get/list all pods in all namespaces).
This would be a separate improvement in addiotion to #1551 (which didn't play out quite as I initially thought).
Why is this needed:
Better management of NRT objects. Potentially fewer stale objects in the cluster.
The text was updated successfully, but these errors were encountered:
What would you like to be added:
We could possibly improve the behavior of nfd-gc by removing stale NRT objects (whose creator pod does not exist anymore). The
ownerReference
field shouldn't be used as that creates warnings (see k8s garbage-collection). We could use an nfd-specific annotation (likenfd.node.kubernetes.io/owner-pod: <namespace>/<pod-uuid>
). The nfd-gc would delete the object if the pod doesn't exist. Possibly acting only if<namespace>
is the same where nfd-gc itself is running (so that it doesn't need cluster-wide access to get/list all pods in all namespaces).This would be a separate improvement in addiotion to #1551 (which didn't play out quite as I initially thought).
Why is this needed:
Better management of NRT objects. Potentially fewer stale objects in the cluster.
The text was updated successfully, but these errors were encountered: