Closed andersy005 closed 4 years ago
@jhamman & @scottyhq
I am trying to use the pangeo-notebook
meta package approach here, however, the notebook
kernel appears to be inaccessible when the notebook is launched:
jovyan@jupyter-andersy005-2dcesm-2dlens-2daws-2dhnb0jm7i:~$ conda env list# conda environments:
#
base /srv/conda
notebook * /srv/conda/envs/notebook
Am I missing something?
Never mind. I just found out that the issue was due missing dependencies on my end.
I am now running into a different issue. KubeCluster()
instantiation with
from dask_kubernetes import KubeCluster
cluster = KubeCluster()
cluster.adapt(minimum=2, maximum=100, wait_count=60)
used to work. However, with the meta package approach, I am getting the following error:
With the pangeo-notebook meta package approach, should I be instantiating KubeCluster()
differently? Cc @jhamman
@andersy005 - the metapackage does not install dask_config.yaml which you still need. Things worked for you previously because we've been including that file in pangeo base images. So
1) you could continue using a base image that has it (see https://github.com/pangeo-data/pangeo-stacks-dev/blob/master/base-image/dask_config.yml). I'm going to try to finalize this pangeo-stacks rework today - here is a binder config that works using the metapackage https://github.com/scottyhq/pangeodev-binder/tree/master/binder.
2) Otherwise you can copy your own dask_config.yaml into the image as is done here https://github.com/pangeo-data/pangeo-cloud-federation/tree/staging/deployments/icesat2/image/binder
Thank you for the clarification, @scottyhq! I previously misunderstood how the meta-package was intended to work. I am going to use the base image option.
Superseded by #51
First stab at using Pangeo-notebook conda meta package.