Skip to content

Conversation

@jacobtomlinson
Copy link
Member

Closes #3084

Instead of waiting for the scheduler connection for a long time we are waiting for a shorter time but retrying until the original timeout. If the connection is never made then we raise.

@mrocklin
Copy link
Member

Any thoughts on how to test this? I can't think of any, but I thought I'd ask.

@jacobtomlinson
Copy link
Member Author

jacobtomlinson commented Sep 30, 2019

It's a tricky one as this is trying to fix a strange edge case.

Mocking the connect to timeout and then succeed would work, but that is very much code testing that the code does what the code does*, rather than actually testing a real example.

Perhaps something fancy with iptables could do it. But I'm not sure if it is worth the effort.

*intentionally unpleasant sentence

@mrocklin
Copy link
Member

Yeah, that was my sense as well. Just thought I'd ask. Happy to merge on passed CI.

@mrocklin mrocklin merged commit 6fe6277 into dask:master Sep 30, 2019
jacobtomlinson added a commit to jacobtomlinson/dask-kubernetes that referenced this pull request Oct 1, 2019
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Wait on single connection or try multiple connections

2 participants