Maybe fix test_worker_waits_for_scheduler #5995
Conversation
Previously this was timing out in some test runs. One *possible* issue is that we started a coroutine to ask the scheduler to release us, but the scheduler doesn't exist. This removes that request. I'm not confident that this will fix the issue. I give around a 30% chance. However, I am fairly confident that this won't have negative consequences.
fjetter
left a comment
fjetter
left a comment
There was a problem hiding this comment.
The report is definitely unnecessary.
| pass | ||
|
|
||
|
|
||
| @pytest.mark.slow |
There was a problem hiding this comment.
I would expect this test run very fast, assuming the timeout in wait_for is reduced. At least after the report is turned off.
There was a problem hiding this comment.
It tests that the worker waits for at least three seconds. I though about removing it but I think that it genuinely adds signal.
|
This should overall accelerate the test but I doubt it will resolve the flakyness. The I traced down timing out workers on close to #5883 and #5910 |
|
Cool. Thank you for the pointer. Looking. |
|
I haven't seen this test being flaky recently (https://dask.org/distributed/test_report.html) |
2 participants
Previously this was timing out in some test runs.
One possible issue is that we started a coroutine to ask the
scheduler to release us, but the scheduler doesn't exist. This
removes that request.
I'm not confident that this will fix the issue. I give around a 30%
chance. However, I am fairly confident that this won't have negative
consequences.