Release 2025.4.0

[jrbourbeau]

Release version	2025.4.0
Planned release date	2025-04-22
Status	On-track
Release manager	@jrbourbeau

Best effort

Try to close before the release but will not block the release

• Scheduler not robust to many connected clients distributed#9043
  • Reduce size of scheduler_info distributed#9045
  • Cache WorkerState host property distributed#9044
  • Dynamically scale heartbeat and scheduler_info intervals distributed#9046

Blocker

Issues that would cause us to block and postpone the release if not fixed

• xarray store disables blockwise fusion if compute=False dask#11879
• ...

Comments

cc @fjetter @hendrikmakait @TomAugspurger @jacobtomlinson

[TomAugspurger]

I think things should be OK from my POV.

I'm working through one more set of failures on rapidsai/dask-upstream-testing#40 related to lost futures. Right now, I'm hopeful those can be resolved in cuml by simplifying how some of the array / dataframe conversion is done, but I'll post an update later today.

[TomAugspurger]

dask/dask#11886 would be nice to get in, if Florian and others are comfortable with the changes there.

[fjetter]

Unless there is a reason to hurry I would suggest to wait for another day or two. The PRs that are in flight right now are pretty core and test coverage isn't as great as I'd like it to be. In the past weeks, merging things in dask often caused regressions in distributed, etc.
I suggest to wait for anohter day or two to see if something else comes up. @TomAugspurger If you could confirm that your CI works once this is all merged, that would be a good signal as well

[TomAugspurger]

In the latest run at https://github.com/rapidsai/dask-upstream-testing/actions/runs/14452894460/job/40529534529#step:10:2411, which had

• dask: dask/dask@8639b6e
• distributed: dask/distributed@6691e27

All the tests passed except for one failing thanks to a bug in cudf. Thanks @fjetter!

The only other failure I'm tracking is at dmlc/xgboost#10994, which is maybe from dask/distributed#8998 though I haven't had luck getting to a root cause. Regardless, that's been around for a while now so shouldn't be considered a blocker for this release.

[fjetter]

Just triggered another CI run in distributed with clean environment cache after merging the dask PR (see https://github.com/dask/distributed/actions/runs/14492717736) if that is not more broken than usual we can move forward

[jrbourbeau]

Just a heads up that I'll start pushing this out in a few minutes

[jrbourbeau]

Closing as 2025.4.0 is out on PyPI, conda-forge, and docker registries