Uploaded image for project: 'OpenShift Bugs'
  1. OpenShift Bugs
  2. OCPBUGS-22095

PerformanceProfile render fails at Day-0 because the master/worker pools are not yet present

XMLWordPrintable

    • Moderate
    • No
    • CNF Compute Sprint 248, CNF Compute Sprint 249, CNF Compute Sprint 250, CNF Compute Sprint 251, CNF Compute Sprint 252, CNF Compute Sprint 253
    • 6
    • False
    • Hide

      None

      Show
      None
    • Hide
      With this release, the PerformanceProfile now renders correctly on Day-0. The installer supports pre-rendering of the PerformanceProfile related manifests. Previously, on single-node OpenShift the Machine Config Operator (MCO) render was executed after the Performance Profile render and so the control plane and worker MachineConfigPools were created too late. This issue has been addressed by ensuring the Node Tuning Operator knows what to do when control plane and worker MachineConfigPools are not yet there.
      Show
      With this release, the PerformanceProfile now renders correctly on Day-0. The installer supports pre-rendering of the PerformanceProfile related manifests. Previously, on single-node OpenShift the Machine Config Operator (MCO) render was executed after the Performance Profile render and so the control plane and worker MachineConfigPools were created too late. This issue has been addressed by ensuring the Node Tuning Operator knows what to do when control plane and worker MachineConfigPools are not yet there.
    • Bug Fix
    • Done
    • 2024-04-10: Needs a retest due to the last comment. The issue here is likely fixed already.

      Description of problem:

      The installer supports pre-rendering of the PerformanceProfile related manifests. However the MCO render is executed after the PerfProfile render and so the master and worker MachineConfigPools are created too late.

This causes the installation process to fail with:

Oct 18 18:05:25 localhost.localdomain bootkube.sh[537963]: I1018 18:05:25.968719       1 render.go:73] Rendering files into: /assets/node-tuning-bootstrap
Oct 18 18:05:26 localhost.localdomain bootkube.sh[537963]: I1018 18:05:26.008421       1 render.go:133] skipping "/assets/manifests/99_feature-gate.yaml" [1] manifest because of unhandled *v1.FeatureGate
Oct 18 18:05:26 localhost.localdomain bootkube.sh[537963]: I1018 18:05:26.013043       1 render.go:133] skipping "/assets/manifests/cluster-dns-02-config.yml" [1] manifest because of unhandled *v1.DNS
Oct 18 18:05:26 localhost.localdomain bootkube.sh[537963]: I1018 18:05:26.021978       1 render.go:133] skipping "/assets/manifests/cluster-ingress-02-config.yml" [1] manifest because of unhandled *v1.Ingress
Oct 18 18:05:26 localhost.localdomain bootkube.sh[537963]: I1018 18:05:26.023016       1 render.go:133] skipping "/assets/manifests/cluster-network-02-config.yml" [1] manifest because of unhandled *v1.Network
Oct 18 18:05:26 localhost.localdomain bootkube.sh[537963]: I1018 18:05:26.023160       1 render.go:133] skipping "/assets/manifests/cluster-proxy-01-config.yaml" [1] manifest because of unhandled *v1.Proxy
Oct 18 18:05:26 localhost.localdomain bootkube.sh[537963]: I1018 18:05:26.023445       1 render.go:133] skipping "/assets/manifests/cluster-scheduler-02-config.yml" [1] manifest because of unhandled *v1.Scheduler
Oct 18 18:05:26 localhost.localdomain bootkube.sh[537963]: I1018 18:05:26.024475       1 render.go:133] skipping "/assets/manifests/cvo-overrides.yaml" [1] manifest because of unhandled *v1.ClusterVersion
Oct 18 18:05:26 localhost.localdomain bootkube.sh[537963]: F1018 18:05:26.037467       1 cmd.go:53] no MCP found that matches performance profile node selector "node-role.kubernetes.io/master="

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

      4.14.0-rc.6

      How reproducible:

      Always

      Steps to Reproduce:

      1. Add an SNO PerformanceProfile to extra manifest in the installer. Node selector should be: "node-role.kubernetes.io/master="
2.
3.

      Actual results:

      no MCP found that matches performance profile node selector "node-role.kubernetes.io/master="

      Expected results:

      Installation completes

      Additional info:

      apiVersion: performance.openshift.io/v2
kind: PerformanceProfile
metadata:
 name: openshift-node-workload-partitioning-sno
spec:
 cpu:
   isolated: 4-X <- must match the topology of the node
   reserved: 0-3
 nodeSelector:
   node-role.kubernetes.io/master: ""

              msivak@redhat.com Martin Sivak
              msivak@redhat.com Martin Sivak
              Dwaine Gonyier Dwaine Gonyier
              Egli Hila
              Votes:
              0 Vote for this issue
              Watchers:
              13 Start watching this issue

                Created:
                Updated:
                Resolved: