Jupyter notebook can't find kernels for conda environments from nb_conda_kernels

Question

I use different conda environments for different Python projects, and used to happily use nb_conda_kernels to access kernels for those environments in the Jupyter notebook. They would be available in the 'New' drop-down with names like Python [conda env:project42]. After some updates, none of these kernels are showing up in Jupyter, and when I try to open a notebook that had been using one of these kernels I get Kernel not found - Could not find a kernel matching Python [conda env:project42] *. Please select a kernel:. The only option is Python 3.

I've checked that all the environments are still there in conda (with conda env list) and that each one has the ipykernel package installed.

The environment where I'm launching Jupyter from has all of the following (among many others):

ipykernel 5.1.4 py37h39e3cac_0 ipython 7.13.0 py37h5ca1d4c_0 jupyter 1.0.0 py37_7 jupyter_client 6.0.0 py_0 nb_conda 2.2.1 py37_0 nb_conda_kernels 2.2.2 py37_0 python 3.7.0 h6e4f718_3

Should I be using jupyter_conda instead of nb_conda? How can I see what nb_conda_kernels is trying to do or what errors it is having? I can create kernels manually but it would be time consuming to do this for every environment I have.

Any tips greatly appreciated! Let me know if I can provide any more version/system details.

Answer 1

I fixed it by upgrading nb_conda_kernels to version 2.2.3! The release notes mention that this version is a fix for jupyter_client 6. I created a new environment just for the notebook so it's possible there were some other factors at play as well. In the new environment I ran conda install --channel conda-forge nb_conda_kernels to install version 2.2.3.