Skip to content

Performance regression in multiple task-based shuffle tests #706

@hendrikmakait

Description

@hendrikmakait

Multiple tests show a performance regression introduced around March 07, 2023:
Screenshot 2023-03-08 at 17 48 52

Screenshot 2023-03-08 at 17 49 35

Screenshot 2023-03-08 at 17 50 02

The regression appears to be in the context of task-based shuffling.

Related CI failures:

Bisecting dask/dask via A/B tests suggests dask/dask#9994 as the cause of this regression.

Metadata

Metadata

Assignees

No one assigned

    Labels

    affected: upstreamThis issue affected the upstream version of coiled-runtimecause: unknownRoot cause could not be identified but issue didn't occur againregression

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions