Release Dates
- FLIP Freeze: (A FLIP must be accepted by this date in order to be considered for this release. Note, any FLIP that may not be implemented in a week, or that might destabilize the release, should be deferred.)
- Feature Freeze: (major features merged & working on stabilization, minor features have PR, release branch cut; anything not in this state will be automatically moved to the next release in JIRA)
- Code Freeze:
The release manager is Dong Lin and Yun Gao .
How to Contribute
Before code freeze:
- Participate in votes and discussions to land or postpone the open FLIPs.
- Review patches. The more reviewers, the more content we can fit in.
- Write unit/integration/system tests.
After code freeze:
- Write more unit/integration/system tests.
- Improve documentation.
- Test the release candidates.
- Open blocker JIRAs on critical issues found. Open non-blocker JIRAs on any other issues found.
- Fix critical bugs.
- Review bug fixes.
- Vote on release candidates. Even though only PMC votes are binding, community votes are super important as we evaluate the readiness of the release.
Highlighted Features
- API and Infrastructure
- Supported stages requiring multi-input multi-output
- Supported online learning with APIs exposing model data
- Improved usability for managing parameters
- Added tools for composing DAG of stages into a new stage
- Added stream-batch Unified Iteration Library
- Algorithm library addition
- logistic regression
- k-means
- k-nearest neighbors
- Naive Bayes
- one-hot encoder
Open Issues
Planned FLIP Content
Note: The planned content is not binding - final content will be based the features committed by branch-cutting date. See Flink Improvement Proposals for the full list of KIPs.