Fix timeout in client.restart

[mdwint]

I noticed client.restart() would hang sometimes, and then I discovered the timeout passed to asyncio.wait_for being expressed as a "deadline" instead. Am I correct in assuming this is wrong?

[jrbourbeau]

Good catch @mdwint! You're correct that timeout in asyncio.wait_for is the number of seconds to wait.

Could you add a small test to distributed/tests/test_client.py to ensure that we're timing out as expected?

Also, I know it predates the changes in this PR, but could you add a call to parse_timedelta so timeout= can accept inputs like "10min", "13s" instead of just an integer or float number of seconds? For instance, here's an example of where we use parse_timedelta in another location in client.py:

distributed/distributed/client.py

Line 605 in be2d495

timeout = parse_timedelta(timeout, "s")

[mdwint]

Thanks for your guidance @jrbourbeau. I've added that test, as well as the call to parse_timedelta. Let me know if this is ok.

[jrbourbeau]

This looks great! Just a heads up, I pushed a small commit to use "0.5s" instead of 0.5 in the test you added to make sure the parse_timedelta addition is working as expected -- hope that's okay. Will merge once CI finishes

[jrbourbeau]

Also I noticed this was your first code contribution to this repository. Welcome!

[mdwint]

Thanks, @jrbourbeau!