Burn down guide


This is a guide for the steps required to perform at the start of burn down.

Given the current development version is v0.A.0, then the next version is v0.B.0 where B=A+1.

Intended audience



  • Create branch-0.B branches
  • Create 0.B project board
  • Create 0.B gpuCI jobs
  • Announce burn down



In each RAPIDS project, create a new branch for 0.B such as:

git fetch origin
git checkout -b branch-0.B origin/branch-0.A`


On this new branch, update documentation and version numbers.

  • README.md - including build badges
  • CHANGELOG.md - Add new section for 0.B.0
  • Files in ci/release/update-version.sh
  • Any other documentation files


Commit, tag, and push the changes:

git commit -am "DOC Update to v0.B"
git tag v0.B.0a
git push origin HEAD:branch-0.B
git push origin v0.B.0a

Project boards

For each RAPIDS project, navigate to the current release board and follow these instructions to copy it as the v0.B board.

gpuCI jobs

Update the gpuCI configuration file to include the branch-0.B. Then run the RAPIDS Seed Job to generate the new jobs.


Update each of the auto-mergers to merge from 0.A to 0.B.

Make sure to update all of the following

  • Job name
  • Job description
  • SCM Branch
  • GH_HEAD & GH_BASE parameters in the downstream job


Announce that burn down has begun and when it will end.

Suggested template:

@channel  :fire::arrow_down: *RAPIDS v0.9 Burn Down Announcement* :fire::arrow_down:

:warning: cuDF/cuML/cuGraph/RMM/cuStrings/dask-cuda v0.9 have moved to the burn down stage - `branch-0.10` is available but *not the default branch yet*

*Burn down ends Tuesday, August 13*
See https://docs.rapids.ai/maintainers for full v0.9 schedule

Please keep the following in mind:
- *Stop adding issues/PRs for v0.9*; unless deemed critical by the PICs
- *Concentrate all dev efforts* to close any issues or PRs on the v0.9 boards
- *Check open PRs* to ensure they target the correct branch before merging
- *Move open issues/PRs* to the new v0.9 boards and branch
- Auto-mergers are in place to merge updates from v0.9 to v0.10 - https://gpuci.gpuopenanalytics.com/view/gpuCI%20-%20auto-mergers/

See https://docs.rapids.ai/releases/process/#burn-down for more details on the burn down and development process.

*v0.9 boards:*
- cuDF - https://github.com/rapidsai/cudf/projects/15
- cuML - https://github.com/rapidsai/cuml/projects/8
- cuGraph - https://github.com/rapidsai/cugraph/projects/7
- RMM - https://github.com/rapidsai/rmm/projects/5
- cuStrings - https://github.com/rapidsai/custrings/projects/4
- dask-cuda - https://github.com/rapidsai/dask-cuda/projects/2

*v0.10 boards:*
- cuDF - https://github.com/rapidsai/cudf/projects/16
- cuML - https://github.com/rapidsai/cuml/projects/10
- cuGraph - https://github.com/rapidsai/cugraph/projects/10
- RMM - https://github.com/rapidsai/rmm/projects/6
- cuStrings - https://github.com/rapidsai/custrings/projects/5
- dask-cuda – https://github.com/rapidsai/dask-cuda/projects/3