Release v0.12.0

[klihub]

Release Process

• Create and push new release branch
• Local release preparations (on the release branch)
  • Run e2e tests
  • Sync/tidy up dependencies.
    • Run go mod tidy.
    • Run git commit -m 'go.mod,go.sum: update dependencies.' go.{mod,sum}, if necessary.
  • Run git tag -a -m "NRI plugins $VERSION" $VERSION.
• Publishing
  • Push the tag with git push $VERSION.
  • Check that new container images are published for the tag.

    • # You can do this with the in-repo artifact verification script:
      ./scripts/release/check-artifacts.sh --images $VERSION
      

  • Finalize the new draft release created by CI
    • Check that all artefacts (Helm charts) were uploaded
    • Write the change log to the release.
    • Get the change log OK'd by other maintainers.
  • Publish the draft as a release.
  • Verify that the Helm repo was updated

    • # You can do this with the in-repo artifact verification script:
      ./scripts/release/check-artifacts.sh --charts $VERSION
      # Or to do a final verification of both images and charts:
      ./scripts/release/check-artifacts.sh $VERSION
      

  • Add a link to the tagged release in this issue.
  • Generate the operator bundle by running make bundle within deployment/operator directory and submit the generated content to the community-operators.
• Create and push unannotated development tag X.Y.0-devel for next release cycle.
• Close this issue.

[askervin]

First draft as basis for release notes:

This new release of NRI Reference Plugins brings both new features and bug fixes to resource policy plugins.

What's New

Resource information passing

• Deploying new resource-annotator to the cluster enables NRI
  plugins to get exact devices and numbers in resource requests and
  limits. This enables proper resource managements of burstable
  containers that request more than 256 CPUs. Kubelet configures
  cgroups CPU shares linearly up to 256 requested CPUs, but after that
  all containers get the same CPU shares, even if they would request
  257, 512 or 1024 CPUs.

Telemetry

• Topology-aware and balloons plugins support OpenTelemetry for
  metrics collection. There is a minor change in Prometheus output in
  debugging-oriented metrics in the balloons policy.

Balloons Policy

• Balancing balloons on balloon types enables creating equally
  many balloons from two or more balloon types for the same
  purpose. For example, if balloon types B0, ..., B3 are local to
  similar but separate hardware resources that accelerate HPC
  containers, this feature enables balancing HPC containers so that
  equally many containers get their CPUs from each of these
  types. Containers are assigned to a composite balloon type BC whose
  components are B0, ..., B3, and componentCreation strategy
  of the BC balloon type is set to balance-balloons.

E2E Test Framework

• CPU and CXL memory hotplug from e2e test scripts.

• Emulating 4096-CPU systems with the limitation of having only a few CPUs plugged at boot.

• Custom kernel building and installing from test scripts from any
  kernel source tree and branch.

[klihub]

v0.12.0 tagged and published: https://github.com/containers/nri-plugins/releases/tag/v0.12.0