Get more informative timeout variable names or clarify them in the comments

[falfaroalmagro]

Could the variable timeout in line 6318 of the scheduler have a more sensible name?

I have no idea how to configure that parameter in the yaml configuration file. I have resorted to manually set that value to 600 seconds in my own copy of the code, because I have not been able to configure it. I have tried:

• jobqueue.slurm.connect-timeout
• jobqueue.slurm.tcp-timeout
• jobqueue.slurm.no-workers-timeout
• jobqueue.slurm.death-timeout
• jobqueue.slurm.timeout

None of those correspond to that timeout variable.

Thanks

[jacobtomlinson]

It looks to me like the scheduler function you are referring to gets called via Client.scatter(), which optionally takes timeout as an argument. The default value for that timeout is set by distributed.comm.timeouts.connect.

distributed/distributed/client.py

Line 1068 in 4840709

timeout = dask.config.get("distributed.comm.timeouts.connect")

I'm not sure renaming an internal variable within a private function makes things any clearer. Perhaps part of the confusion is the docstring mentions that this is the timeout for the initial connection to the scheduler, but that default value seems to be used to other timeouts in other places too, like scatter.

distributed/distributed/client.py

Lines 968 to 969 in 4840709

	timeout: int (defaults to configuration ``distributed.comm.timeouts.connect``)
	Timeout duration for initial connection to the scheduler