NOTE
This project was migrated to Eclipse Gitlab on October 14, 2021.
The Eclipse Foundation toolkit for maintaining permissions across multiple version-control platforms. This includes scripts for syncronizing user permissions from PMI to Github or Gitlab, running automated backups of repositories, and managing bot access through a variety of means.
To be granted access to a repository through the Eclipse Sync scripts (Gitlab or Github), a user must first either be nominated as a committer or a project lead within the PMI(projects management infrastructure), or be added as a contributor by an active committer or project lead. Depending on the role granted within the PMI, different access rights will be granted through the sync scripts. Should a user be promoted or retired within a project, the new permission sets should be active within a few hours of finalization.
Bot access within repositories is possible, but managed by a manual process and tracked by a publicly available API rather than through the sync script. How these bot permissions are typically interpreted varies by platform, and more info for each is available in the Github and Gitlab Sync sections within this readme.
The Eclipse Foundation supports granting permissions from triage to maintain permissions on Github, and Reporter to Maintainer on Gitlab. Owner permissions are not supported for either platform as they are not needed for typical project management scenarios.
Within Github, there is a mixed strategy for management of projects within the space. Projects that are started while under the Eclipse umbrella or from a project that was incepted within the Eclipse ecosystem are by default created under the central Eclipse organization. On request, projects can be migrated to a separate organization that is still managed by the EclipseWebmaster account. Repositories or projects born from organizations or groups that have joined Eclipse Foundation post inception are usually managed under organizations managed by the EclipseWebmaster. While there are cases where projects can cross organizational bounds, it is uncommon (and covered by the sync script).
Permissions to projects are managed through hidden teams that are then granted access to each repository for the given project within the current organization. For each organization that a project has repositories in, a set of contributor, committer, and project-lead teams will be created to give access to those repositories. Each of these teams will have the same set of users as defined within the project management interface on projects.eclipse.org. For users to be properly added by this mechanism, they must set their Github Handle within their Eclipse Account on accounts.eclipse.org.
In regards to bot access, this is typically granted at the repository level, but can also be added at the team level if more broad access is needed. These permissions, while not removed by the script are currently managed manually by the Eclipse Foundation. If there are issues regarding bot access, new or existing, an issue should be created within our bug-tracking system rather than within this project.
Below is an example of how a few projects may be managed within the Eclipse ecosystem:
Eclipse/
├─ dash-licenses (project)
├─ technology-dash-contributors (private team, access to dash-licenses)
├─ technology-dash-committers (private team, access to dash-licenses)
├─ technology-dash-committers (private team, access to dash-licenses)
locationtech/
├─ spatial4j (project)
├─ locationtech-spatial4j-contributors (private team, access to spatial4j)
├─ locationtech-spatial4j-committers (private team, access to spatial4j)
├─ locationtech-spatial4j-committers (private team, access to spatial4j)Information on Github permissions is available in the documentation for organizations.
| Eclipse Group | Github Permission |
|---|---|
| Contributor | triage |
| Committer | push |
| Project Leads | maintain |
In Gitlab, a nested group strategy was chosen to manage access to both groups and projects. This gives greater control over inherited permissions without having to manage teams across multiple base groups. For each Open Source group with repositories managed by the Eclipse Foundation (such as Eclipse Foundation and the OpenHWGroup), a base group will exist to encapsulate all projects for that group. Within each of these groups, each active project will have a subgroup (such as Eclipse Dash and Eclipse Marketplace Client) that will manage permissions for all repositories active within the Gitlab instance.
In regards to bot access, this can be granted at either the subgroup or project (repository) level depending on the needs of the project. These permissions, while not removed by the script are currently managed manually by the Eclipse Foundation. If there are issues regarding bot access, new or existing, an issue should be created within our bug-tracking system rather than within this project.
Below is an example of a few projects within the Eclipse Gitlab instance and their structure:
Eclipse/ (group)
├─ Eclipse Dash/ (group)
│ ├─ dash-gitlab-testing (project)
│ ├─ org.eclipse.dash.handbook (project)
├─ Eclipse Marketplace Client/ (group)
│ ├─ MPC Client (project)
│ ├─ org.eclipse.epp.mpc (project)
Eclipse Foundation/ (group)
├─ webdev/ (group)
│ ├─ eclipsefdn-api-common (project)More information on Gitlab permissions can be found in the API documentation.
| Eclipse Group | Gitlab Permission |
|---|---|
| Contributor | Reporter |
| Committer | Developer |
| Project Leads | Maintainer |
The following parameters can be used when running the sync scripts manually.
| Parameter name | Required | Accepts | Default | Description |
|---|---|---|---|---|
| -s, --secretLocation | ✓ | string | N/A | The location of the files containing API access tokens and secure configurations containing keys. |
| -c, --console | x | boolean flag | false |
Adds console output to logging statements for the current run |
| -d, --dryRun | x | boolean flag | false |
Runs script in dry run mode, not writing any changes to the API. |
| -D, --deletionDryRun | x | boolean flag | false |
Runs script in deletion dry run mode, not performing any deletion operations to the API. |
| -t, --devMode | x | boolean flag | false |
Runs the script with the dev mode active. This changes the returned data from the Eclipse API to affect a dev sandbox rather than production projects. |
| -p, --project | x | string | N/A | The project ID (e.g. technology.dash) of the project that should be updated (at the exclusion of all other projects) |
| -h, --help | x | N/A (flag) | N/A | Prints the help text for the script parameters. |
| -V, --verbose | x | boolean flag | false |
Sets the script to run in verbose mode (ranges from 1-4 for more verbose logging). |
| Parameter name | Required | Accepts | Default | Description |
|---|---|---|---|---|
| -s, --secretLocation | ✓ | string | N/A | The location of the files containing API access tokens and secure configurations containing keys. |
| -c, --console | x | boolean flag | false |
Adds console output to logging statements for the current run |
| -d, --dryRun | x | boolean flag | false |
Runs script as dry run, not writing any changes to API. |
| -D, --devMode | x | boolean flag | false |
Runs the script with the dev mode active. This changes the returned data from the Eclipse API to affect a dev sandbox rather than production projects. |
| -H, --host | x | string | https://gitlab.eclipse.org |
The Gitlab host target for the sync. This allows for testing and staging migrations for use in development and dry runs. |
| -h, --help | x | N/A (flag) | N/A | Prints the help text for the script parameters. |
| -p, --provider | x | string | oauth2_generic |
The OAuth provider name set in GitLab for the Eclipse Accounts binding. |
| -P, --project | x | string | N/A | The project ID (e.g. technology.dash) of the project that should be updated (at the exclusion of all other projects) |
| -V, --verbose | x | boolean flag | false |
Sets the script to run in verbose mode (ranges from 1-4 for more verbose logging). |
By default, the script is run in docker containers to emulate the production environment (Openshift). This sync tool can be run in standard and verbose mode. The difference between the modes is that in verbose all log messages are printed to the STDOUT of the container.
Before running, an api-token file should be created that contains the GitHub API key to be used for connectivity. This should be created in a secret folder in the root of the project (this has been excluded from Git commits so there is less danger of pushing live keys to a public repository).
Additionally, an eclipse-oauth-config should be created when running the GitLab sync script. This file will define how connections to the Eclipse OAuth server should be handled. If this is missing, the GitLab sync script will fail. The file should be in JSON format, with the following being an example of format:
{"oauth":{"timeout":3600, "client_id":"<client id>","client_secret":"<client secret>","endpoint":"https://accounts.eclipse.org","scope":"eclipsefdn_view_all_profiles","redirect":"http://localhost"}}docker build -f Dockerfile -t ef/gh-test .
docker run -i --rm -v <fullpath to current project folder>/secrets:/run/secrets --env DRYRUN=true ef/gh-testOnce a release candidate has been identified for production, tests should be run locally using the master branch to ensure that the integration of patches are stable as the pull requests that had been validated. Additionally, npm ci && npm run test should be run to ensure that the current package lock file is stable, passes tests, and contains no live vulnerabilities from upstream packages. If there is issues with this, patches should be created to address these issues before the release goes live.
When determining the version bump semantic versioning spec should be used, which also forms the base of how NPM versioning works. The new version should be applied through a PR on the master branch. Once this is done a new tag can be created on the master branch with the format v#.#.#, where the numbers should match the version stated in the package.json file. While there is no anticipated need to create an NPM package release, it is good practice and can help keep code clean.
Once the master branch tag is created it can then be merged into the production branch. Pushing code to this branch triggers a build to update the running containers that manage this script. The golive process should be coordinated with the Releng team to ensure that support and additional eyes are available to help validate the changes. If there are issues regarding the new version, the script should be rolled back to the previous version to avoid service interruptions via pushes to the production branch.
Martin Lowe (Eclipse Foundation)
Copyright 2019 the Eclipse Foundation, Inc. and the eclipsefdn-github-sync authors. Code released under the Eclipse Public License Version 2.0 (EPL-2.0).