Uploaded image for project: 'OpenShift Bugs'
  1. OpenShift Bugs
  2. OCPBUGS-63455

VPA recommendation dropped to minimal request due to upstream issue #7726

XMLWordPrintable

    • Icon: Bug Bug
    • Resolution: Done
    • Icon: Normal Normal
    • 4.19.z
    • 4.16
    • Pod Autoscaler
    • Quality / Stability / Reliability
    • False
    • Hide

      None

      Show
      None
    • 3
    • Moderate
    • None
    • None
    • None
    • AUTOSCALE - Sprint 280
    • 1
    • Proposed
    • Bug Fix
    • Hide
      Before this update, rounding errors occurred when the Vertical Pod Autoscaler (VPA) Recommender loaded histograms from checkpoints following a restart. This caused the VPA Recommender to produce invalid memory recommendations with a value of 0, which could lead to pods receiving insufficient resources and failing to perform correctly. With this release, the logic for loading histogram data has been corrected to eliminate these rounding errors. As a result, the VPA Recommender now consistently provides accurate memory recommendations even after a restart, ensuring the stability and proper scaling of managed workloads. (link:https://issues.redhat.com/browse/OCPBUGS-63455[OCPBUGS-63455])
      Show
      Before this update, rounding errors occurred when the Vertical Pod Autoscaler (VPA) Recommender loaded histograms from checkpoints following a restart. This caused the VPA Recommender to produce invalid memory recommendations with a value of 0, which could lead to pods receiving insufficient resources and failing to perform correctly. With this release, the logic for loading histogram data has been corrected to eliminate these rounding errors. As a result, the VPA Recommender now consistently provides accurate memory recommendations even after a restart, ensuring the stability and proper scaling of managed workloads. (link: https://issues.redhat.com/browse/OCPBUGS-63455 [ OCPBUGS-63455 ])
    • None
    • None
    • None
    • None

      Description of problem:

      VPA and encountered an issue where the VPA recommendation dropped to minimal request, preventing pods from starting. This seems to align with a known upstream issue[1]

As per above, the issue has been fixed in vpa-release 1.3 and newer, which is included in Kubernetes 1.33. 

However, our latest OpenShift version 4.19 is still based on Kubernetes 1.32.
      I am raising this jira ticket to backport the fix to previous version of OpenShift as per https://redhat-internal.slack.com/archives/C02F1J9UJJD/p1761046027851589
      - [1] https://github.com/kubernetes/autoscaler/issues/7726

              rh-ee-macao Max Cao
              rhn-support-harspati Harshada Patil
              Paul Rozehnal
              None
              Paul Rozehnal Paul Rozehnal
              None
              Votes:
              0 Vote for this issue
              Watchers:
              7 Start watching this issue

                Created:
                Updated:
                Resolved: