wait_for_workers got stuck when to create cluster but application failed on yarn

[FANLONGFANLONG]

I put 500 GB for a worker accidentally to start yarn cluster. But looks likely

1. I can create cluster(application) on yarn because i got cluster.app_id and cluster.scheduler_address.

2. i use client.wait_for_workers to wait for worker ready

3. but dask got stuck.
   

4. i checked application on yarn and it is failed already
   

the issue is reproduced.

May I know what I could do to avoid the problem?

[fjetter]

This sounds like a yarn specific problem and I suggest to repost this on the issue tracker of https://github.com/dask/dask-yarn

[jrbourbeau]

Just transferred this issue over to the dask-yarn repo

[FANLONGFANLONG]

@jrbourbeau thanks