The Multus Daemonset does not configure a priorityClassName like other CNI's, which can cause the Multus Pod to be preempted in favor of less important Pods
What did you expect to happen?
Multus Pods should configure a priorityClassName, preferable system-node-critical like other CNI's
How can we reproduce it (as minimally and precisely as possible)?
The command kubectl get pods -A -o json | jq '.items[] | select (.spec.priorityClassName == null) | "Namespace: " + .metadata.namespace + " Name: " + .metadata.name' which lists Pods that do not have a priorityClassName returns "multus" pods.
OS
not important
Version of Ansible
not important
Version of Python
not important
Version of Kubespray (commit)
24dc4cef5
Network plugin used
calico
Full inventory with variables
not important
Command used to invoke ansible
not important
Output of ansible run
not important
Anything else we need to know
I will create a pull-request to add priorityClassName support for Multus
What happened?
The Multus Daemonset does not configure a priorityClassName like other CNI's, which can cause the Multus Pod to be preempted in favor of less important Pods
What did you expect to happen?
Multus Pods should configure a priorityClassName, preferable
system-node-critical
like other CNI'sHow can we reproduce it (as minimally and precisely as possible)?
The command
kubectl get pods -A -o json | jq '.items[] | select (.spec.priorityClassName == null) | "Namespace: " + .metadata.namespace + " Name: " + .metadata.name'
which lists Pods that do not have a priorityClassName returns "multus" pods.OS
not important
Version of Ansible
not important
Version of Python
not important
Version of Kubespray (commit)
24dc4cef5
Network plugin used
calico
Full inventory with variables
not important
Command used to invoke ansible
not important
Output of ansible run
not important
Anything else we need to know
I will create a pull-request to add priorityClassName support for Multus