Can we raise a more informative error when papermill is not installed?

[krassowski]

Context

When papermill is not in environment the job fails without any sensible description. There are some ideas to solve it with LBYL (nebari-dev/nebari#1838):

argo-jupyter-scheduler/argo_jupyter_scheduler/utils.py

Lines 135 to 137 in 1d9f1d1

	def gen_conda_env_path(conda_env_name: str, use_conda_store_env: bool = True):
	# TODO: validate that `papermill` is in the specified conda environment

but I think what would be neat is if we could return an informative error from the container itself:

currently it just shows "Workflow failed"

Maybe it is as simple as adjusting these lines:

argo-jupyter-scheduler/argo_jupyter_scheduler/utils.py

Lines 207 to 210 in 1d9f1d1

	papermill = f"papermill -k {kernel_name} {input_path} {output_path}"
	jupyter = f"jupyter nbconvert --to html {output_path} --output {html_path}"
	return f'conda run -p {conda_env_path} /bin/sh -c "{{ {papermill} && {jupyter} ; }} >> {log_path} 2>&1"'

?

Value and/or benefit

Less user confusion. Users using the scheduler function rather than giving up after first attempt because it is not working.

Anything else?

No response

[Adam-D-Lewis]

Definitely agree with this 👍

[nkaretnikov]

I've implemented this in #10.