Uploaded image for project: 'Red Hat OpenShift AI Engineering'
  1. Red Hat OpenShift AI Engineering
  2. RHOAIENG-6276

[2.8.z] Infinite reconciliation loop of DataScienceCluster (DSC)

XMLWordPrintable

    • Testable

      What

      DSC keeps reconciling the components. It seems happening when codeflare and/or kueue and/or ray are enabled

      UPDATE: from latest tests on RHOAI 2.8.1 and 2.8.0, I think it happens when kueue and ray are enabled together

      Steps to reproduce

      1. install RHOAI 2.8.1 or 2.9.0 (we haven't checked previous releases, but it could be present even before)
      2. create DSC enaling kueue and ray (or all of the components)
      3. wait until all the components get deployed in redhat-ods-applications namespace
      4. check DSC: still reconciling
      5. disable the above components
      6. wait for components pod to be removed
      7. check DSC: not reconciling anymore

      Additional info

      Environment

      • IBM Cloud with RHOAI 2.9 pre-RC iib:704539 and OCP 4.15.8
      • PSI Cluster (OpenStack) with RHOAI 2.8.1 from OperatorHub (stable channel) and OCP 4.15.8

            astefanu@redhat.com Antonin Stefanutti
            rhn-support-bdattoma Berto D'Attoma
            Abhijeet Dhumal Abhijeet Dhumal
            Votes:
            0 Vote for this issue
            Watchers:
            7 Start watching this issue

              Created:
              Updated:
              Resolved: