Automate releases on Github and PyPi.
This is a bot that helps maintainers deliver their software to users. It is meant to watch github repositories for
release pull requests. The PR must be named in this format 0.1.0 release
. No other format is supported yet.
Once the PR is merged, bot will create a new Github release, PyPi and Fedora respectively.
Changelog will be pulled from root of the
repository and must be named CHANGELOG.md
. Changelog for the new
version must begin with version heading, i.e # 0.1.0
.
Everything between this heading and the heading for previous version will be pulled into the changelog.
Alternatively, you can let the bot do the boring work, update __version__
variable and fill changelog with commit messages from git log.
You can trigger this action by creating an issue and name it the same as you would a release PR, e.g. 0.1.0 release
.
All you have to do after that is merge the PR that the bot will make.
The bot works with
pypa/setuptools_scm plugin. If
you're using it, you don't need to care about __version__
at all. You can be
also sure that the bot will make the PyPI release correctly — before it
releases the software, it checks out the tag in the git repo.
A release-conf.yaml
file is required. See Configuration section for details.
Once a Github release is complete, bot will upload this release to PyPI. Note that you have to setup your login details (see Requirements).
After PyPI release, if enabled in release-conf.yaml
,
bot will try to release on Fedora dist-git, on master
branch and branches specified in configuration.
It should not create merge conflicts, but in case it does,
you have to solve them first before attempting the release again.
$ pip install release-bot
Other possible installations are through Docker, OpenShift.
First interaction with release bot may be automated releases on Github. Let's do it.
This is meant to be upstream repository where new releases will be published.
Within upstream repository create release-conf.yaml
file which contains info on how to release the specific project.
Copy and edit release-conf.yaml.
At the end of release-conf.yaml
add this line of code:
# whether to allow bot to make PRs based on issues
trigger_on_issue: true
For possible advanced setup check the documentation for an upstream repository.
Create configuration file conf.yaml
. You can use one from this repository. You will need to generate a Github personal access token.
Recommended permissions for access token are: repo
, delete_repo
, user
.
At the end of conf.yaml
add this line of code:
# Name of the account that the github_token belongs to
# Only needed for triggering the bot on an issue.
github_username: <your_github_username>
Note: This file should not be stored in upstream repository as it contains sensitive data.
For possible advanced setup check the documentation for a private repository. Also, see requirements in case you want include PyPi releases.
At this point, release-bot is installed. At least two configuration files are set release-conf.yaml
and conf.yaml
(optionally .pypirc
).
Launch bot by a command:
$ release-bot -c <path_to_conf.yaml> --debug
You can scroll down and see debug information of running bot.
- Create an issue having
0.0.1 release
as a title in your upstream repository. You can select your own version numbers. - Wait for the bot to make a new PR based on this issue (refresh interval is set in
conf.yaml
). - Once the PR is merged bot will make a new release.
- Check release page of your upstream repository at GitHub and you should see new release
0.0.1
.
Since now, feel free to create releases automatically just by creating issues.
There are two yaml configuration files:
conf.yaml
-- a config for the bot itself with some sensitive data (recommended to store in private repo)release-conf.yaml
-- stored in upstream repository and contains info on how to release the specific project.
You need to setup a git repository, where you'll store the conf.yaml
and .pypirc
files.
If you are releasing on Fedora, you will also need to add id_rsa
(a private ssh key that you configured in FAS) and fedora.keytab
(kerberos keytab for fedora).
If this is not a local repository, make sure it's private so you prevent any private info leaking out.
If the path to conf.yaml
is not passed to bot with -c/--configuration
,
bot will try to find it in current working directory.
Here are the conf.yaml
configuration options:
Option | Description | Required |
---|---|---|
repository_name |
Name of your Github repository | Yes |
repository_owner |
Owner of the repository | Yes |
github_token |
Github personal access token | Yes |
github_username |
Name of the account that the github_token belongs to. Only needed for triggering the bot on an issue. |
No |
github_app_installation_id |
Installation ID (a number) of the Github app. | No |
github_app_id |
ID (a number) of the Github app. | No |
github_app_cert_path |
Path to a certificate which Github provides as an auth mechanism for Github apps. | No |
fas_username |
FAS username. Only need for releasing on Fedora | No |
refresh_interval |
Time in seconds between checks on repository. Default is 180 | No |
Sample config named conf.yaml can be found in this repository.
Regarding github_token
, it's usually a good idea to create a Github account for the bot
(and use its Github API token)
so you can keep track of what changes were made by bot and what are your own.
You can also create a Github app and use it as an authentication mechanism for
the bot. For that you need to specify the three config values prefixed with
github_app
.
You also have to have a release-conf.yaml
file in the root of your upstream project repository.
Here are possible options:
Option | Meaning | Required |
---|---|---|
changelog |
List of changelog entries. If empty, changelog defaults to $version release |
No |
author_name |
Author name for changelog. If not set, author of the merge commit is used | No |
author_email |
Author email for changelog. If not set, author of the merge commit is used | No |
fedora |
Whether to release on fedora. False by default | No |
fedora_branches |
List of branches that you want to release on. Master is always implied | No |
trigger_on_issue |
Whether to allow bot to make PRs based on issues. False by default. | No |
labels |
List of labels that bot will put on issues and PRs | No |
Sample config named release-conf-example.yaml can be found in this repository.
Are specified in requirements.txt
.
You have to setup your PyPI login details in $HOME/.pypirc
as described in
PyPI documentation.
If you are releasing to Fedora, you will need to have an active kerberos ticket while the bot runs
or specify path to kerberos keytab file with -k/--keytab
.
Also, fedpkg
requires that you have ssh key in your keyring, that you uploaded to FAS.
To make it easier to run this, release-bot is available as an source-to-image builder image.
You can then create the final image like this:
$ s2i build $CONFIGURATION_REPOSITORY_URL usercont/release-bot app-name
where $CONFIGURATION_REPOSITORY_URL is link to repository with conf.yaml and .pypirc files.
To test it locally, you can the run the final image like this:
$ docker run <app-name>
once all changes, configuration files exist in GitHub and git repository contains needed files, you can try to create an issue in your GitHub repository with string like "X.Y.Z release" and you can see log like this:
$ docker run meta-test-family-bot
---> Setting up ssh key...
Agent pid 12
Identity added: ./.ssh/id_rsa (./.ssh/id_rsa)
11:47:36.212 configuration.py DEBUG Loaded configuration for fedora-modularity/meta-test-family
11:47:36.212 releasebot.py INFO release-bot v0.4.1 reporting for duty!
11:47:36.212 github.py DEBUG Fetching release-conf.yaml
11:47:51.636 releasebot.py DEBUG No merged release PR found
11:47:52.196 releasebot.py INFO Found new release issue with version: 0.8.4
11:47:55.578 releasebot.py DEBUG No more open issues found
11:47:56.098 releasebot.py INFO Making a new PR for release of version 0.8.5 based on an issue.
11:47:57.608 utils.py DEBUG ['git', 'clone', 'https://github.com/fedora-modularity/meta-test-family.git', '.']
...
You can also run this bot in OpenShift using openshift-template.yml in this repository.
You must set two environment variables, the $APP_NAME
is the name of your release-bot deployment,
and $CONFIGURATION_REPOSITORY
which contains configuration for the release-bot.
The contents of the repository are described above.
Note that if you use private repository (which you absolutely should),
you will need to set up a new OpenShift secret named
release-bot-secret
to authenticate. It can be a ssh private key that you can use to access the repository
(for GitHub see deploy keys).
Here's an guide on
how to do that in OpenShift GUI, or another
guide
that uses oc
commandline tool.
By default, the release-bot builder image won't update itself when a
new version of this image is pushed to docker hub.
You can change it by uncommenting lines with #importPolicy:
and #scheduled: true
in openshift-template.yml.
Then the image will be pulled on a new release.
If you are interested in making contribution to release-bot project, please read Contribution guide for more information.