Uploaded image for project: 'Knative Serving'
  1. Knative Serving
  2. SRVKS-1021

[FLAKE] TestAutoscaleSustaining/aggregation-weighted-exponential

XMLWordPrintable

    • Icon: Bug Bug
    • Resolution: Done
    • Icon: Undefined Undefined
    • 1.28.0
    • 1.26.0
    • None
    • None

      The test currently fails on OCP 4.13 randomly
      FAIL: https://master-jenkins-csb-serverless-qe.apps.ocp-c1.prod.psi.redhat.com/job/functional_tests/job/live/91/
      PASS: https://master-jenkins-csb-serverless-qe.apps.ocp-c1.prod.psi.redhat.com/job/functional_tests/job/live/90/

      Error message:

      autoscale_test.go:119: interim scale didn't fulfill constraints: revision "autoscale-sustaining-aggregatio-cwzcsibq-00001" #replicas: 8, want at least: 9
      

      I was not able to spot any issues in controller or autoscaler logs. Also, no issues with insufficient CPU/Memory on OpenShift side.

      We run the test on a node with 2 cCPUs and the test needs to spin up 10 concurrent go routines that send requests and make Serving scale the ksvc to 10 pods. This is different from upstream where they have 16 vCPUs and midstream where we run in a container without CPU limits.
      The theory is that a node with higher vCPUs should make the test pass.

      There's also upstream issue https://github.com/knative/serving/issues/1304 which has as problem running Kind.

      Slack discussion: https://redhat-internal.slack.com/archives/CF5ANN61F/p1675761655265219

              mgencur@redhat.com Martin Gencur
              mgencur@redhat.com Martin Gencur
              Votes:
              0 Vote for this issue
              Watchers:
              2 Start watching this issue

                Created:
                Updated:
                Resolved: