CAPZ minor versions (that is, 1.5.0 versus 1.4.x) are released every two months.
CAPZ patch versions (for example, 1.5.2 versus 1.5.1) are released as often as weekly. Each week at the open office hours meeting, maintainers decide whether or not a patch release is called for based on community input. A patch release may bypass this cadence if circumstances warrant.
The two most recent minor releases of CAPZ will be supported with bug fixes. Since minor releases arrive every two months, each minor release receives fixes for four months.
For example, let's assume CAPZ v1.4.2 is the current release, and v1.3.2 is the latest in the previous minor release line. When v1.5.0 is released, it becomes the current release. v1.4.2 becomes the previous release line and remains supported. And v1.3.2 reaches end-of-life and no longer receives support through bug fixes.
Note that "support" in this context refers strictly to whether or not bug fixes are backported to a release line. Please see the support documentation for more general information about how to get help with CAPZ.
Any significant user-facing bug fix that lands in the main
branch should be backported to the current and previous release lines. Security-related fixes are automatically considered significant and user-facing.
Improvements or significant changes to tests should be backported to the current release line. This is intended to minimize friction in the event of a critical test fix. Test improvements or changes may sometimes need to be backported to the previous release line in the event that tests break on all release branches.
Experimental Cluster API features (for example, AzureManagedCluster
) may evolve more rapidly than graduated v1 features. CAPZ allows general changes, enhancements, or additions in this area to be cherry-picked into the current release branch for inclusion in patch releases. This will accelerate the effort to graduate experimental features to the stable API by allowing faster adoption and iteration.
Breaking changes are also allowed in experimental APIs; those changes will not be included in a patch release, but will be introduced in a new minor release, with appropriate release notes.
Sometimes pull requests touch a large number of files and are more likely to create challenges for the automated cherry-pick process. In such cases, maintainers may prefer to delay merging such changes until the end of a minor release cycle.
- Make sure the metadata.yaml file is up to date and contains the new release with the correct cluster-api contract version.
- If not, open a PR to add it.
This must be done prior to generating release artifacts, so the release contains the correct metadata information for clusterctl to use.
- Create a new GitHub milestone for the next release
- Change milestone applier so new changes can be applied to the appropriate release
- Open a PR in https://github.com/kubernetes/test-infra to change this line
- Example PR: kubernetes/test-infra#16827
- Open a PR in https://github.com/kubernetes/test-infra to change this line
This can be done in parallel with release publishing and does not impact the release or its artifacts.
Using that same next release version used to create a new milestone, update the the capz provider metadata.yaml that we use to run PR and periodic cluster E2E tests against the main branch templates.
For example, if the latest stable API version of capz that we run E2E tests against is v1beta
, and we're releasing v1.12.0
, and our next release version is v1.13.0
, then we want to ensure that the metadata.yaml
defines a contract between v1.13.0
and v1beta1
:
apiVersion: clusterctl.cluster.x-k8s.io/v1alpha3
releaseSeries:
- major: 1
minor: 11
contract: v1beta1
- major: 1
minor: 12
contract: v1beta1
- major: 1
minor: 13
contract: v1beta1
Additionally, we need to update the type: InfrastructureProvider
spec in azure-dev.yaml to express that our intent is to test (using the above example) 1.5
. By convention we use a sentinel patch version "99" to express "any patch version". In this example we want to look for the type: InfrastructureProvider
with a name
value of v1.12.99
and update it to v1.13.99
:
- name: v1.13.99 # "vNext"; use manifests from local source files
Update the API version upgrade tests to use the oldest supported release versions of CAPI and CAPZ after the release is cut as "Init" provider versions. See this PR for more details.
-
If you don't have a GitHub token, create one by going to your GitHub settings, in Personal access tokens. Make sure you give the token the
repo
scope. -
Fetch the latest changes from upstream and checkout the main branch:
git fetch upstream git checkout main
-
Generate release notes by running the following command:
export GITHUB_TOKEN=<your GH token> export RELEASE_TAG=v1.2.3 # change this to the tag of the release to be cut make release-notes
-
Review the release notes file generated at
CHANGELOG/<RELEASE_TAG>.md
and make any necessary changes. -
Open a pull request with the release notes.
Note: Important! The commit should only contain the release notes file, nothing else, otherwise automation will not work.
Merging the PR will automatically trigger a Github Action to create a release branch (if needed), push a tag, and publish a draft release.
- Images are built by the post push images job. This will push the image to a staging repository.
- Wait for the above job to complete for the tag commit and for the image to exist in the staging directory, then create a PR to promote the image and tag:
make promote-images
This will automatically create a PR in k8s.io and assign the CAPZ maintainers. Example PR: kubernetes/k8s.io#4284.
For reviewers of the above-created PR, to confirm that the resultant image SHA-to-tag addition is valid, you can check against the staging repository.
Using the above example PR, to verify that the image identified by SHA d0636fad7f4ced58b5385615a53b7cb2053f79c4788bd299e0ac9e46a25b5053
has the expected v1.4.3
, tag, you would inspect the image metadata by viewing it in the Google Container Registry UI:
- Manually format and categorize the release notes
- Ensure that the promoted release image is live. For example:
docker pull registry.k8s.io/cluster-api-azure/cluster-api-azure-controller:${RELEASE_TAG}
- Publish release
- Announce the release
cluster-api-provider-azure follows the semantic versionining specification.
Example versions:
- Pre-release:
v0.1.1-alpha.1
- Minor release:
v0.1.0
- Patch release:
v0.1.1
- Major release:
v1.0.0
- A release yaml file
infrastructure-components.yaml
containing the resources needed to deploy to Kubernetes - A
cluster-templates.yaml
for each supported flavor - A
metadata.yaml
which maps release series to cluster-api contract version - Release notes
For major and minor releases we will need to update the set of capz-dependent test-infra
jobs so that they use our latest release branch. For example, if we cut a new 1.3.0
minor release, from a newly created release-1.3
git branch, then we need to update all test jobs to use capz at release-1.3
instead of release-1.2
.
Here is a reference PR that applied the required test job changes following the 1.3.0
minor release described above:
Go to the Netlify branches and deploy contexts in site settings and click "edit settings". Update the "Production branch" to the new release branch and click "Save". The, go to the Netlify site deploys and trigger a new deploy.
Note: this step requires access to the Netlify site. If you don't have access, please ask a maintainer to update the branch.
Consider whether anything should be updated in the roadmap document by answering the following questions:
- Have any of the Epics listed been entirely or largely achieved? If so, then the Epic should likely be removed and highlighted during the release communications.
- Are there any new Epics we want to highlight? If so, then consider opening a PR to add them and bringing them up in the next office hours planning meeting with the milestone review.
- Have any updates to the roadmap document occurred in the past 6 months? If not, it should be updated in some form.
If any changes need to be made, it should not block the release itself.
- Announce the release in Kubernetes Slack on the #cluster-api-azure channel.
- Follow the communications process for pre-releases
- An announcement email is sent to
kubernetes-sig-azure@googlegroups.com
andkubernetes-sig-cluster-lifecycle@googlegroups.com
with the subject[ANNOUNCE] cluster-api-provider-azure <version> has been released