Uploaded image for project: 'OpenShift Bugs'
  1. OpenShift Bugs
  2. OCPBUGS-19480

CNV regression with recent Kubernetes rebase - device plugin

XMLWordPrintable

    • Critical
    • No
    • Rejected
    • False
    • Hide

      None

      Show
      None
    • Hide
      20230919: GREEN OCP team likely to pull patch version from septembers (including the fixes) but final decision still pending (capacity, other prios)
      20230911: GREEN started conversation about the timeline to pull the upcoming kube patch releases
      20230905: GREEN 1.27, 1.26, 1.25 backports all merged u/s and on track for the september patch update. Will sync with OCP folks to pull the fixes
      20230829: 1.27 and 1.26 backports approved, pending decision of u/s release team for september patch release. 1.25 not approved yet.
      8/21: working u/s CI issues in prep for Sep bump
      Show
      20230919: GREEN OCP team likely to pull patch version from septembers (including the fixes) but final decision still pending (capacity, other prios) 20230911: GREEN started conversation about the timeline to pull the upcoming kube patch releases 20230905: GREEN 1.27, 1.26, 1.25 backports all merged u/s and on track for the september patch update. Will sync with OCP folks to pull the fixes 20230829: 1.27 and 1.26 backports approved, pending decision of u/s release team for september patch release. 1.25 not approved yet. 8/21: working u/s CI issues in prep for Sep bump

      Description of problem:

      Pods are being terminated on Kubelet restart if they consume any device.

In case of CNV this Pods are carrying VMs and the assuption is that Kubelet will not terminate the Pod in this case.

      Version-Release number of selected component (if applicable):

      4.14 / 4.13.z / 4.12.z

      How reproducible:

      This should be reproducable with any device plugin as far as goes my understanding

      Steps to Reproduce:

      1. Create Pod requesting device plugin
2. Restart Kubelet
3.

      Actual results:

      Admission error -> Pod terminates

      Expected results:

      No error -> Existing & Running Pods will continue running after Kubelet restart

      Additional info:

      The culprit seems to be https://github.com/kubernetes/kubernetes/pull/116376

            fromani@redhat.com Francesco Romani
            lpivarc Luboslav Pivarc
            Sunil Choudhary Sunil Choudhary
            Votes:
            0 Vote for this issue
            Watchers:
            6 Start watching this issue

              Created:
              Updated:
              Resolved: