Uploaded image for project: 'Network Observability'
  1. Network Observability
  2. NETOBSERV-1012

Add PriorityClassName in operator + Loki

XMLWordPrintable

    • Icon: Story Story
    • Resolution: Unresolved
    • Icon: Undefined Undefined
    • None
    • None
    • Operator
    • 3
    • False
    • None
    • False

      We should define different priorityClassName per components (agent / FLP / plugin) and unique cluster wide to allow admins to declare PriorityClass and decide how Netobserv components will be prioritized.

      https://kubernetes.io/docs/concepts/scheduling-eviction/pod-priority-preemption/

      https://redhat-internal.slack.com/archives/C03PUMXPZGQ/p1683018287734889?thread_ts=1682335591.430679&cid=C03PUMXPZGQ

      https://github.com/operator-framework/operator-lifecycle-manager/blob/master/doc/design/adding-priority-classes.md

       

      /!\ If the priority class is not found, the Pod is rejected

       

      This is set on logging collector: https://github.com/openshift/cluster-logging-operator/blob/b9c22742f6ffeb3e1c7eb76091176a85ccee9387/internal/collector/collector.go#L22 however it's not set on Loki side since it runs on dedicated nodes.

            Unassigned Unassigned
            jpinsonn@redhat.com Julien Pinsonneau
            Votes:
            0 Vote for this issue
            Watchers:
            2 Start watching this issue

              Created:
              Updated: