Application unittests using utils_test.cluster() timeout on Windows

[mpeleshenko]

My application's distiributed unittests always seem to fail on Windows due to the timeout set in the below line. They work fine on Linux. I tried changing it to 15s on Windows and it worked fine. Can this timeout be increased or made configurable for cases where it takes longer to create a worker process?

distributed/distributed/utils_test.py

Line 653 in 538767b

worker['address'] = worker['queue'].get(timeout=5)

Code:

scheduler_kwargs = {
    'preload': (<path to module>,),
    'preload_argv': (<arg for dask_setup>,),
}
worker_kwargs = {
    'preload': (<path to module>,),
    'preload_argv': (<arg for dask_setup>,),
    'ncores': 1,
    'memory_limit': 70000000000,
}
with distributed.utils_test.cluster(nworkers=1, nanny=True, worker_kwargs=worker_kwargs, scheduler_kwargs=scheduler_kwargs) as (s, ws):
    <run something on cluster>

Exception:

Traceback (most recent call last):
  File "<conda env>\lib\site-packages\distributed\utils_test.py", line 653, in cluster
    worker['address'] = worker['queue'].get(timeout=5)
  File "<conda env>\lib\multiprocessing\queues.py", line 105, in get
    raise Empty
queue.Empty

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
...
  File "<conda env>\lib\contextlib.py", line 59, in __enter__
    return next(self.gen)
  File "<conda env>\lib\site-packages\distributed\utils_test.py", line 655, in cluster
    raise pytest.xfail.Exception("Worker failed to start in test")
_pytest.outcomes.XFailed: Worker failed to start in test

Python 3.5.6
distributed 1.25.1

[mrocklin]

Yes, the timeout could be increased. I'd be happy to merge a PR with this change. (sorry for the late response)