Issue Policy

The MLflow Issue Policy outlines the categories of MLflow GitHub issues and discusses the guidelines & processes associated with each type of issue.

Before filing an issue, make sure to search for related issues and check if they address yours.

For support (ex. "How do I do X?"), please ask on Stack Overflow.

Issue Categories

Our policy is that GitHub issues fall into one of the following categories:

1. Feature Requests
2. Bug reports
3. Documentation fixes
4. Installation issues

Each category has its own GitHub issue template. Please do not delete the issue template unless you are certain your issue is outside its scope.

Feature Requests

Guidelines

Feature requests that are likely to be accepted:

• Are minimal in scope (note that it's always easier to add additional functionality later than remove functionality)
• Are extensible (e.g. if adding an integration with an ML framework, is it possible to add similar integrations with other frameworks?)
• Have user impact & value that justifies the maintenance burden of supporting the feature moving forwards. The JQuery contributor guide has an excellent discussion on this.

Lifecycle

Feature requests typically go through the following lifecycle:

1. A feature request GitHub Issue is submitted, which contains a high-level description of the proposal and its motivation. We encourage requesters to provide an overview of the feature's implementation as well, if possible.
2. The issue is triaged to identify whether more information is needed from the author, give an indication of priority, and route feature requests to appropriate committers.
3. The feature request is discussed with a committer. The committer will provide input on the implementation overview or ask for a more detailed design, if applicable.
4. After discussion & agreement on the feature request and its implementation, an implementation owner is identified.
5. The implementation owner begins developing the feature and ultimately files associated pull requests against the MLflow Repository or packages the feature as an MLflow Plugin.

Bug reports

Guidelines

In order to ensure that maintainers are able to assist in any reported bug:

• Ensure that the bug report template is filled out in its entirety with appropriate levels of detail, particularly in the Code to reproduce issue section.
• Verify that the bug you are reporting meets one of the following criteria:
  • A recent release of MLflow does not support the operation you are doing that an earlier release did (a regression).
  • A documented feature or functionality does not work properly by executing a provided example from the docs.
  • Any exception raised is directly from MLflow and is not the result of an underlying package's exception (e.g., don't file an issue that MLflow can't log a model that can't be trained due to a tensorflow Exception)
• Make a best effort to diagnose and troubleshoot the issue prior to filing.
• Verify that the environment that you're experiencing the bug in is supported as defined in the docs.
• Validate that MLflow supports the functionality that you're having an issue with. A lack of a feature does not constitute a bug.
• Read the docs on the feature for the issue that you're reporting. If you're certain that you're following documented guidelines, please file a bug report.

Bug reports typically go through the following lifecycle:

1. A bug report GitHub Issue is submitted, which contains a high-level description of the bug and information required to reproduce it.
2. The bug report is triaged to identify whether more information is needed from the author, give an indication of priority, and route to request appropriate committers.
3. An MLflow committer reproduces the bug and provides feedback about how to implement a fix.
4. After an approach has been agreed upon, an owner for the fix is identified. MLflow committers may choose to adopt ownership of severe bugs to ensure a timely fix.
5. The fix owner begins implementing the fix and ultimately files associated pull requests.

Documentation fixes

Documentation issues typically go through the following lifecycle:

1. A documentation GitHub Issue is submitted, which contains a description of the issue and its location(s) in the MLflow documentation.
2. The issue is triaged to identify whether more information is needed from the author, give an indication of priority, and route the request to appropriate committers.
3. An MLflow committer confirms the documentation issue and provides feedback about how to implement a fix.
4. After an approach has been agreed upon, an owner for the fix is identified. MLflow committers may choose to adopt ownership of severe documentation issues to ensure a timely fix.
5. The fix owner begins implementing the fix and ultimately files associated pull requests.

Installation issues

Installation issues typically go through the following lifecycle:

1. An installation GitHub Issue is submitted, which contains a description of the issue and the platforms its affects.
2. The issue is triaged to identify whether more information is needed from the author, give an indication of priority, and route the issue to appropriate committers.
3. An MLflow committer confirms the installation issue and provides feedback about how to implement a fix.
4. After an approach has been agreed upon, an owner for the fix is identified. MLflow committers may choose to adopt ownership of severe installation issues to ensure a timely fix.
5. The fix owner begins implementing the fix and ultimately files associated pull requests.