Bokeh warning from get_task_stream.

/home/nfs/an.taugspurger/miniconda3/envs/ucx-dev-test/lib/python3.7/site-packages/bokeh/io/saving.py:126: UserWarning: save() called but no resources were supplied and output_file(...) was never called, defaulting to resources.CDN
  warn("save() called but no resources were supplied and output_file(...) was never called, defaulting to resources.CDN")
         Took 14.14s

To avoid the warning, we need to pass a resources= to https://github.com/dask/distributed/blob/fb30c33562862f30864456766424b44a3e91aa5b/distributed/client.py#L3590

The bokeh default of resources=resources.CDN seems good, but I don’t think we should warn.

I could imagine making this configurable, for when you want to view the bokeh plot behind a firewall that doesn’t have access to bokeh’s CDN. To make it configureable, we’ll need to add a resoruces keyword to Client.get_task_stream and distributed.get_task_stream. Is there a potential ambiguity with the name resources? I could imagine “resources”? http://distributed.dask.org/en/latest/resources.html?highlight=resources#worker-resources Maybe call it bokeh_resources to be safe.

[jrbourbeau]

Great, thanks @prasunanand. I think we want to have a call to output_file before save, like is done here:

https://github.com/dask/distributed/blob/b0c000883eb10d4801b967fe348c4d6281ca3f1d/distributed/scheduler.py#L5097-L5098

[prasunanand]

I am working on it.