Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Adapt to non-string task keys in distributed #1225

Merged
merged 2 commits into from
Aug 25, 2023

Conversation

wence-
Copy link
Contributor

@wence- wence- commented Aug 25, 2023

Now that keys are no longer strings there are two places we must adapt here.

  1. Explicit comms must no longer manually stringify task keys before staging and intersection with the on-worker data (since that data mapping doesn't use the stringified version)
  2. The zict.File-backed slow buffer in DeviceHostFile needs to translate non-string keys to string keys before writing to disk, to do this, use the same implementation that distributed uses for its own spilling buffer.

Now that task keys are not always strings, the file-backed zict
mapping needs to explicitly stringify keys. Distributed has such a
mapping implemented so we just import it and use it here.
Now that task keys are no longer strings when stored in the worker
data dictionary, we should not stringify the set of keys when staging.

- Closes rapidsai#1224
@wence- wence- requested a review from a team as a code owner August 25, 2023 14:38
@github-actions github-actions bot added the python python code needed label Aug 25, 2023
@wence- wence- added bug Something isn't working non-breaking Non-breaking change labels Aug 25, 2023
Copy link
Member

@pentschev pentschev left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Awesome, thanks so much for working on this @wence- ! 😄

@pentschev
Copy link
Member

/merge

@rapids-bot rapids-bot bot merged commit 390ad36 into rapidsai:branch-23.10 Aug 25, 2023
19 checks passed
@wence- wence- deleted the wence/fix/1224 branch August 25, 2023 18:50
@charlesbluca
Copy link
Member

Once the next release is out, we will probably want to bump our minimum Dask pinning to 2023.8.2, as this PR now makes it such that we fail to import with the latest stable release (2023.8.1)

@pentschev
Copy link
Member

pentschev commented Aug 31, 2023

Once the next release is out, we will probably want to bump our minimum Dask pinning to 2023.8.2, as this PR now makes it such that we fail to import with the latest stable release (2023.8.1)

I wouldn't mind this but we generally agreed to only do this at release time, the reason for that is there's a heavy load to do that across all projects. Once we have the metapackage to pin Dask versions across RAPIDS this will become much simpler.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
bug Something isn't working non-breaking Non-breaking change python python code needed
Projects
None yet
Development

Successfully merging this pull request may close these issues.

Tests failing after Distributed upstream changes
3 participants