Skip to content

map_overlap tasks fail to deserialize on workers - keywords must be strings #6624

@gjoseph92

Description

@gjoseph92
import distributed
import numpy as np
import dask.array as da


if __name__ == "__main__":
    v = da.random.random((20, 20), chunks=(5, 5))

    overlapped = da.map_overlap(np.sum, v, depth=2, boundary="reflect")

    client = distributed.Client()
    overlapped.compute()
2022-06-23 19:16:55,060 - distributed.core - ERROR - keywords must be strings
Traceback (most recent call last):
  File "/Users/gabe/dev/distributed/distributed/core.py", line 849, in handle_stream
    handler(**merge(extra, msg))
  File "/Users/gabe/dev/distributed/distributed/worker.py", line 1818, in _
    event = cls(**kwargs)
  File "<string>", line 13, in __init__
  File "/Users/gabe/dev/distributed/distributed/worker_state_machine.py", line 667, in __post_init__
    self.run_spec = SerializedTask(**self.run_spec)  # type: ignore[unreachable]
TypeError: keywords must be strings

...

2022-06-23 19:16:55,068 - distributed.nanny - ERROR - Worker process died unexpectedly
    self._target(*self._args, **self._kwargs)
  File "/Users/gabe/dev/distributed/distributed/nanny.py", line 846, in watch_stop_q
    child_stop_q.close()
  File "/Users/gabe/miniconda3/envs/dask-distributed/lib/python3.9/multiprocessing/queues.py", line 143, in close
    self._reader.close()
  File "/Users/gabe/miniconda3/envs/dask-distributed/lib/python3.9/multiprocessing/connection.py", line 182, in close
    self._close()
  File "/Users/gabe/miniconda3/envs/dask-distributed/lib/python3.9/multiprocessing/connection.py", line 366, in _close
    _close(self._handle)

...

distributed.scheduler.KilledWorker: ("('random_sample-concatenate-1086077ac09ace0ac4330fce33825511', 2, 3)", <WorkerState 'tcp://127.0.0.1:65387', name: 2, status: closed, memory: 0, processing: 32>)

cc @crusaderky @fjetter

Environment:

  • Dask version: dc019ed
  • Python version: 3.9.5
  • Operating System: macOS
  • Install method (conda, pip, source): source

Metadata

Metadata

Assignees

Labels

Type

No type

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions