Stream: python-questions

Topic: NCARCLuster errors on cheyenne

view this post on Zulip Brian Bonnlander (Sep 03 2020 at 21:46):

I need a little guidance on how to run Dask properly on cheyenne. I have been doing most of my development on Casper until now, and it's my understanding I have to switch from using SLURMCluster to NCARCluster when working on cheyenne. But I get errors:

import dask
from ncar_jobqueue import NCARCluster
from distributed import Client

cluster = NCARCluster()

~/miniconda3/envs/lens-conversion/lib/python3.8/site-packages/dask/config.py in get(key, default, config)
    452         k = canonical_name(k, result)
    453         try:
--> 454             result = result[k]
    455         except (TypeError, IndexError, KeyError):
    456             if default is not no_default:

KeyError: 'death-timeout'

I am using a conda environment that I built through miniconda3 on Casper. I have no idea if it works to use such an environment on cheyenne, but I had the feeling it should be OK. Anyone see what I should fix? Thanks!

view this post on Zulip Deepak Cherian (Sep 03 2020 at 22:15):

ugh i have this stupid issue too and it has persisted even after recereating my environment. It's really weird. The code is supposed to catch KeyError but is apparently not doing that!

view this post on Zulip Deepak Cherian (Sep 03 2020 at 22:16):

My solution is to import distributed after creating the cluster. I thought this was some weirdness with my env but this makes me thing there's some more fundamental issue

view this post on Zulip Brian Bonnlander (Sep 03 2020 at 22:23):

Thanks Deepak! That worked for me.

view this post on Zulip Notification Bot (Jan 26 2022 at 19:45):

This topic was moved here from #jupyterlab-hub > NCARCLuster errors on cheyenne jupyterhub by Anderson Banihirwe

Last updated: May 16 2025 at 17:14 UTC