Closed lmanan closed 4 years ago
Hey @MLbyML ! I'll be able to help you further tomorrow (it's the middle of the night here) but one thing that sticks out is that your home path is incorrect, it's missing a leading slash:
home/lalit/singularity
-> /home/lalit/singularity
But actually, if you are using Singularity on sherlock that wouldn't be the path - it should just point to "singularity" which is at /usr/bin/singularity
. You are making a lot of assumptions about the user site being on the path, and it's also not clear to me how you are running this (interactively? from the container? what container?) so what I need from you to debug this further is full information to reproduce your case.
Thanks! Be back in a few hours.
I think it's likely those path errors, but another sanity check you can do is to interact with jupyter to list environments. Here is how I re-created your test case on sherlock (but with fixed paths)
mkdir -p /home/users/vsochat/.local/share/jupyter/kernels/mykernel
# Checking my home and singularity locations
echo $HOME
which singularity
# load jupyter (but are you using one in a container?
module load py-jupyter/1.0.0_py36
# list available kernels
$ jupyter kernelspec list
Available kernels:
python3 /share/software/user/open/py-jupyter/1.0.0_py36/lib/python3.6/site-packages/ipykernel/resources
mykernel /home/users/vsochat/.local/share/jupyter/kernels/mykernel
# the jupyter I'm using
$ which jupyter
/share/software/user/open/py-jupyter/1.0.0_py36/bin/jupyter
At this point you'd want to open an interactive (command line) test terminal using it. But I'll stop here because I don't know how to reproduce what you are doing exactly, let me know. I suspect the permissions error just stems from having the wrong path (since the kernels are loaded when you start it up) but it might be something else.
Hi @vsoch Sorry my post and question were misplaced (I was using a singularity container for running some jobs on my university cluster, but this wasn't related to sherlock). I finally did solve my issue though: thank you very much for the quick replies!
Hello,
I am trying to use the python3 installed in the container (
which python
givesopt/conda/bin/python
) to show up as a kernel in jupyter.To do so, in the location
${HOME}/.local/share/jupyter/kernels
, I specify a new directory calledmykernel
and added akernel.json
file of the following description:But this gives a kernel not found error (shown below) upon starting a jupyter notebook. I think the paths to the
singularity
andtest.simg
are correct and this error stems from some permissions issue. Any clue on how to solve this? Thanks.