Uploaded image for project: 'OpenShift Bugs'
  1. OpenShift Bugs
  2. OCPBUGS-42805

In OCP baremetal deployments kube-api is constantly flooded with Node List requests from keepalived.

XMLWordPrintable

    • Quality / Stability / Reliability
    • False
    • Hide

      None

      Show
      None
    • None
    • None
    • None
    • None
    • None
    • None
    • None
    • None
    • None
    • None
    • None
    • None
    • None

      Description of problem:

      In baremetal deployments the kube-api is constantly using CPU as it has to reply to node list requests from all nodes.

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

      All?  I saw this first on 4.14 but I assume it's there since day one.

      How reproducible:

      See steps.

      Steps to Reproduce:

      Deploy an OCP baremetal cluster with a few hundred nodes.
Observe kube-api cpu usage.

      Actual results:

      kube-api is constantly busy replying to node list requests from keepalived pods.

      Expected results:

      There shouldn't be periodic node list requests from all nodes.

      Additional info:

      Originally reported here: https://redhat-internal.slack.com/archives/CDCP2LA9L/p1699020802384189
      Dan Williams had a WIP fix here: https://github.com/openshift/baremetal-runtimecfg/pull/285

              mkowalsk@redhat.com Mat Kowalski
              dceara@redhat.com Dumitru Ceara
              None
              None
              Ross Brattain Ross Brattain
              None
              Votes:
              0 Vote for this issue
              Watchers:
              6 Start watching this issue

                Created:
                Updated: