Ryan Mukherjee

I tried to come up with something to reproduce but haven't been able to. We've also been seeing several other network-related/comms issues, so it's possible that our specific workload and...

Sounds good, thanks! I did something like this: ```python try: fut = cluster.client.compute() dask.distributed.wait(fut, timeout=600) except dask.distributed.TimeoutError as curr_exception: error_text = f'{curr_exception}'[:100] #sometimes the error messages are crazy long print(f'task...

Like this it seems: ``` options = gateway.cluster_options() options['environment'] = { 'DASK_DISTRIBUTED__SCHEDULER__WORKER_TTL': '100s', } cluster = gateway.new_cluster(options) ```

I'm not sure. The `DASK_DISTRIBUTED__` environment variables should get passed through to the scheduler and workers when created via the gateway, whereas the `adaptive_period` parameter you mention seems to only...

I also don't see any labels. It seems that the code is looking for files ending in _labels.pkl, but there are no such files contained within the Google Drive link...

Looks good to me. Thank you!

@martindurant what do you think?

> I think you are right in your argument. However, I have a feeling that f.size is never async actually, so it won't matter. open_async/AsyncStreamedFiles could use work! https://github.com/fsspec/s3fs/pull/742 makes...

> Funny, because S3 is the only one where for sure we should know the size before starting the download: the content-size header is always populated and, possibly outside compressive...

I was able to use the [dependency walker](http://www.dependencywalker.com/) to open the .pyd files within the site-packages directory referenced in my error message. Doing so revealing that OpenNI2.dll was missing. I...