Uploaded image for project: 'Knative Serving'
  1. Knative Serving
  2. SRVKS-1286

Allow to override the activator's terminationGracePeriodSeconds

XMLWordPrintable

    • Icon: Bug Bug
    • Resolution: Done
    • Icon: Major Major
    • 1.35.0, 1.34.1
    • 1.34.0, 1.35.0
    • None
    • None
    • False
    • None
    • False
    • Hide
      Previously, long running requests against Knative Service could be terminated due to Activator component being scaled down by HorizontalPodAutoscaler.
      This issue is now fixed. The `terminationGracePeriodSeconds` is set automatically based on the `max-revision-timeout-seconds` setting for Knative Revisions.
      Show
      Previously, long running requests against Knative Service could be terminated due to Activator component being scaled down by HorizontalPodAutoscaler. This issue is now fixed. The `terminationGracePeriodSeconds` is set automatically based on the `max-revision-timeout-seconds` setting for Knative Revisions.

      In docs https://docs.openshift.com/serverless/1.34/knative-serving/config-applications/configuring-revision-timeouts.html we mention that:

      If this value is increased, the activator `terminationGracePeriodSeconds` should also be increased to prevent in-flight requests being disrupted.

       
      However, we don't mention how to set/override the default value which is 600secs.
      This is critical as activator has an hpa deployed and during scaling down pod termination may cause connection disruption.

              skontopo@redhat.com Stavros Kontopoulos
              skontopo@redhat.com Stavros Kontopoulos
              Votes:
              0 Vote for this issue
              Watchers:
              3 Start watching this issue

                Created:
                Updated:
                Resolved: