The release process for the python client involves creating (or updating) a
release branch, updating release tags, and creating distribution packages and
uploading them to pypi.
There are several releases per version:
- snapshot
- a1 (alpha release)
- b1 (beta release)
- final release
Between each release, there is a waiting period of about two weeks for users to
report issues. Typically, there is a single alpha or beta release, but if there
are a higher than expected number of issues there can be multiple releases
(e.g, a2 or b2).
Automated release
1. Run the release script and send a PR
Generate a Github personal access token following instruction
link
export GITHUB_TOKEN=t # github-personal-access-token
export MINOR_VERSION=x
export PATCH_VERSION=y # The latest patch version for the minor version. Not required for snapshot.
To create a snapshot:
$ KUBERNETES_BRANCH=release-1.${MINOR_VERSION} CLIENT_VERSION=${MINOR_VERSION}.0.0+snapshot DEVELOPMENT_STATUS="3 - Alpha" scripts/release.sh
To create an a1 release:
$ KUBERNETES_BRANCH=release-1.${MINOR_VERSION} CLIENT_VERSION=${MINOR_VERSION}.${PATCH_VERSION}.0a1 DEVELOPMENT_STATUS="3 - Alpha" scripts/release.sh
To create a b1 release:
$ KUBERNETES_BRANCH=release-1.${MINOR_VERSION} CLIENT_VERSION=${MINOR_VERSION}.${PATCH_VERSION}.0b1 DEVELOPMENT_STATUS="4 - Beta" scripts/release.sh
To create a stable release:
$ KUBERNETES_BRANCH=release-1.${MINOR_VERSION} CLIENT_VERSION=${MINOR_VERSION}.${PATCH_VERSION}.0 DEVELOPMENT_STATUS="5 - Production/Stable" scripts/release.sh
Checkout the generated local branch (named “automated-release-of-xxx”) to
continue with the remaining steps.
2. README (not required for snapshots)
Update the compatibility matrix and maintenance status in the README file.
3. Submit pull request
For snapshots, create a PR against the master repo.
For actual releases, create:
- a PR against the release branch
- a second PR against the master branch to cherrypick the CHANGELOG and README
changes.
4. (Repo admin) Create release branch
After merging a new snapshot, create a release branch from the master branch.
(Deprecated) Manual release
1. Create or update release branch
The release branch name should have release-x.x format. All minor and pre-releases
should be on the same branch. To update an existing branch with master (only for
latest pre-release):
export RELEASE_BRANCH=release-x.y
git checkout $RELEASE_BRANCH
git fetch upstream
git rebase upstream/$RELEASE_BRANCH
git pull -X theirs upstream master
You may need to fix some conflicts. For auto-generated files, you can commit
either version. They will be updated to the current version in the next step.
2. Update release tags
Release tags are in the “scripts/constants.py” file. These are the constants you
may need to update:
CLIENT_VERSION: Client version should follow x.y.zDn where x,y,z are version
numbers (integers) and D is one of “a” for alpha or “b” for beta and n is the
pre-release number. For a final release, the “Dn” part should be omitted.
Examples:
- 1.0.0a1 (alpha release)
- 2.0.1b2 (beta release)
- 1.5.1 (final release)
DEVELOPMENT_STATUS: Update it to one of the values of “Development Status” in
this list.
After changing constants to proper versions, update the client using this
command:
scripts/update-client.sh
NOTE: If you see a lot of new or modified files under the kubernetes/test/
directory, delete everything except kubernetes/test/test_api_client.py and
kubernetes/test/test_configuration.py.
Commit changes (should be only version number changes) to the release branch.
Name the commit something like “Update version constants for XXX release”.
After you finished the steps above, refer to the section, “Hot issues”, and
apply the manual fixes.
git push upstream $RELEASE_BRANCH
3. Hot issues
Use the scripts/apply-hotfixes.sh script to apply the fixes below in one step.
As mentioned above, the script should be run after finishing the section “Update release tags”. Also, ensure a clean working directory before applying the script.
Commit the manual changes like this PR does.
There are some hot issues with the client generation that require manual fixes.
The steps below are deprecated and only exist for documentation purposess. They should be performed using the scripts/apply-hotfixes.sh script mentioned above.
Restore custom object patch behavior. You should apply this commit
to ensure custom object patch behavior is backwards compatible. For more
details, see #866 and
#959.
Add alias package kubernetes.client.apis with deprecation warning. You need
to add this file
under kubernetes/client/apis/ to ensure the package is backwards compatible.
For more details, see #974
Add ability to the client to be used as Context Manager kubernetes-client/python#1073
Remove the tests directory (ref: commit/ec9c944f076999543cd2122aff2d86f969d82548). See the upstream issue for more information.
Add tests for the default Configuration behavior (ref: pull/1303 and pull/1285). The commit 1ffa61d0650e4c93e0d7f0becd2c54797eafd407 should be cherry-picked.
4. CHANGELOG
Make sure the change logs are up to date here.
If they are not, follow commits added after the last release and update/commit
the change logs to master.
Then based on the release, follow one of next two steps.
5. README
Update the compatibility matrix and maintenance status in the README file.
Submit pull request
Typically after the you’ve completed steps 2-6 above you can push your changes
open a pull request against kubernetes-client:release-x.y
Patch a release branch
If you are releasing a patch to an existing stable release, you should do a
cherry pick first:
scripts/cherry_pick_pull.sh
Do not merge master into a stable release branch. Run the script without
parameters and follow its instructions to create a cherry pick PR. Get the
PR merged then update your local branch:
export RELEASE_BRANCH=release-x.y
git checkout $RELEASE_BRANCH
git fetch upstream
git rebase upstream/$RELEASE_BRANCH
Sanity check generated client
We need to make sure there are no API changes after running update client
scripts. Such changes should be committed to the master branch first. Run this
command:
scripts/update-client.sh
And make sure there is no API change (version number changes should be fine
as they will be updated in the next step anyway). Do not commit any changes at
this step and go back to the master branch if there are any API changes.
Make distribution packages
First make sure you are using a clean version of python. Use virtualenv and
pyenv packages. Make sure you are using python 3.9.1. I would normally do this
on a clean machine:
(install pyenv)
(install pip)
(install virtualenv)
git clean -xdf
pyenv install -s 3.9.1
pyenv global 3.9.1
virtualenv .release
source .release/bin/activate
python --version # Make sure you get Python 3.9.1
pip install twine
Now we need to create a “~/.pypirc” with this content:
[distutils]
index-servers=pypi
[pypi]
repository = https://upload.pypi.org/legacy/
username = kubernetes
TODO: we should be able to pass these parameters to twine directly. My first attempt failed.
Now that the environment is ready, lets create distribution packages:
python setup.py sdist
python setup.py bdist_wheel --universal
ls dist/
You should see two files in dist folder: “kubernetes*.whl” and “kubernetes*.tar.gz”.
TODO: We need a dry-run option and some way to test the package upload process to pypi.
If everything looks good, run this command to upload packages to pypi:
twine upload dist/*
Create Github release
Create a github release by starting from
this page.
Click the Draft a new release button. Name the tag the same as CLIENT_VERSION. Change
the target branch to “release-x.y”. If the release is a pre-release, check the
This is a pre-release option.
Announcement
Send an announcement email to dev@kubernetes.io with the subject: [ANNOUNCE] kubernetes python-client $VERSION is released
Cleanup
deactivate
rm -rf .release
ref: https://packaging.python.org/distributing/
