Jupyter Erroneous Environments Showing up in Conda tab with `nb_conda` extension

Question

Not sure how to fix this, or how it got corrupted. This is conda on a jupyterhub install.

Error Message:

EnvironmentLocationNotFound: Not a conda environment

Pics Below:

Image from Conda tab (there is an redundant anaconda3 env showing up):

Image of error when loading:

Answer 1

This is a confirmed bug in nb_conda for conda version >=4.4:

The reason is the change of conda info --json output. Fresh installation of version 4.3.33 gives envs=[], version >=4.4 gives envs=[root_env_dir] (always non empty).

It's not fixed in current released version yet (nb_conda 2.2.1), but we can fix this bug manually, by changing the current bug code base in your local machine: ~/anaconda3/pkgs/nb_conda-2.2.1-py36h349edbb_0/lib/python3.6/site-packages/nb_conda/envmanager.py, note that your own anaconda root dir (~/anaconda3) and pkgs build version(py36h349edbb_0) might be different.

FROM:

    return {
        "environments": [root_env] + [get_info(env)
                                      for env in info['envs']]
    }

TO:

    return {
        "environments": [root_env] + [get_info(env) for env in info['envs']
                                      if env != root_env['dir']]
    }

Then, shut down any jupyter notebook server and restart one, the anaconda3 env will gone.

Answer 2

This is either a bug in nbconda, ot a stale kernelspec file. to fix it check whether /home/user/.conda/envs/anaconda3 exists, if it does remove it.

Issue a jupyter kernelspec list in a terminal, for me it gives:

$ jupyter kernelspec list
Available kernels:
  code_scattering    /usr/local/share/jupyter/kernels/code_scattering
  meetup             /usr/local/share/jupyter/kernels/meetup
  octave             /usr/local/share/jupyter/kernels/octave
  python2            /usr/local/share/jupyter/kernels/python2
  python3            /usr/local/share/jupyter/kernels/python3
  pythonroot         /usr/local/share/jupyter/kernels/pythonroot

it should list all kernelspec and their location. Find the corresponding env and delete the folder.