Unable to configure S3 content manager on recent jupyter-server versions

[djuarezg]

As for https://discourse.jupyter.org/t/bad-config-encountered-during-initialization-the-contents-manager-class-trait-of-a-singleuserlabapp-instance-expected-a-subclass-of-jupyter-server-services-contents-manager-contentsmanager-not-the-s3contentsmanager-s3contentsmanager/17304, I cannot switch to s3contents as the manager due to getting the same error as above.

My image is based on https://hub.docker.com/layers/jupyter/scipy-notebook/python-3.10.10

import s3contents
 
c = get_config()
c.S3ContentsManager.bucket = "test-bucket-ignore"
c.S3ContentsManager.endpoint_url = "asdasda"
c.S3ContentsManager.access_key_id = "asdasd"
c.S3ContentsManager.secret_access_key = "asdasda"
c.S3ContentsManager.skip_tls_verify = True
c.NotebookApp.contents_manager_class = s3contents.S3ContentsManager

I am getting:

Bad config encountered during initialization: The ‘contents_manager_class’ trait of a SingleUserLabApp instance expected a subclass of ‘jupyter_server.services.contents.manager.ContentsManager’, not the S3ContentsManager S3ContentsManager.

Using the following package versions:

jupyter_core                      5.3.0
jupyter-events                    0.6.3
jupyter-resource-usage            0.7.1
jupyter_server                    2.5.0
jupyter_server_fileid             0.9.0
jupyter-server-mathjax            0.2.6
jupyter-server-proxy              3.2.2
jupyter_server_terminals          0.4.4
jupyter_server_ydoc               0.8.0
jupyter-telemetry                 0.1.0
jupyter-ydoc                      0.2.3
jupyterhub                        3.1.1
jupyterlab                        3.6.3
jupyterlab-code-formatter         1.6.0
jupyterlab-git                    0.41.0
jupyterlab-pygments               0.2.2
jupyterlab_server                 2.22.1
jupyterlab-system-monitor         0.8.0
jupyterlab-topbar                 0.6.1
jupyterlab-vimrc                  0.5.2
jupyterlab-widgets 

Running the following processes on the spawned container:

(base) jovyan@example:~$ ps auxfww
USER         PID %CPU %MEM    VSZ   RSS TTY      STAT START   TIME COMMAND
jovyan     92652  0.0  0.0   7620  4376 pts/0    Ss   11:18   0:00 bash
jovyan     92895  0.0  0.0  10040  1600 pts/0    R+   11:21   0:00  \_ ps auxfww
jovyan         1  0.0  0.0   2784   992 ?        Ss   May01   0:02 tini -g -- start-notebook.sh --ip=127.0.0.1 --port=57307 --notebook-dir=/home/jovyan/work --SingleUserNotebookApp.default_url=/lab --debug
jovyan         7  0.2  0.3 1152852 200016 ?      Sl   May01   2:22 /opt/conda/bin/python3.10 /opt/conda/bin/jupyterhub-singleuser --ip=127.0.0.1 --port=57307 --notebook-dir=/home/jovyan/work --SingleUserNotebookApp.default_url=/lab --debug
jovyan       125  0.0  0.1 1000708 105652 ?      Ssl  May01   0:06  \_ /opt/conda/bin/python -m ipykernel_launcher --profile=python -f /home/jovyan/.local/share/jupyter/runtime/kernel-00d5e0e5-28cb-4574-b25f-544ae0670add.json

PS: Note that I had to apply #142 suggestion in order to make it work.

[djuarezg]

I suspect this is the main reason: jupyter/notebook#5957 (comment)

That got reverted on posterior versions and this package is not ready for that yet.

[danielfrg]

Could be yeah. I dont have much time to work on it these days so if you find time to make a PR and update the test matrix it should be easy to test.

[djuarezg]

@danielfrg not so proficient in these jupyter world but I found a way to make it work. It turns out that when using jupyter_server, the following lines are the reason why the above class mismatch: https://github.com/danielfrg/s3contents/blob/e44f57054a079e1caee9be94c7a0fbf4abd087d2/s3contents/ipycompat.py#L22-25

If you change them like this it works:

-from notebook.services.contents.filecheckpoints import GenericFileCheckpoints
-from notebook.services.contents.filemanager import FileContentsManager
-from notebook.services.contents.manager import ContentsManager
+from jupyter_server.services.contents.filecheckpoints import GenericFileCheckpoints
+from jupyter_server.services.contents.filemanager import FileContentsManager
+from jupyter_server.services.contents.manager import ContentsManager

Same applies to HybridContents if used as well, imports have to come from jupyter_server

Do you have any recommendation on how to make this less dirty?

[danielfrg]

This looks fine to me. If we just wrap that with an if that checks the version of notebook it should work. Its pretty common to do that.

[djuarezg]

Another thing I saw was https://github.com/jupyter/nbdime/blob/c82362344e596efdc4f54c927d90338940e0fa41/nbdime/webapp/nb_server_extension.py#L16-L33, which is a bit dirtier

[danielfrg]

Yes, its not the cleanest but its pretty common to deal with this situations.

[xfu83]

@danielfrg could you help cut a patch release and publish it on Pypi, thanks!

[danielfrg]

Just released 0.10.0. Let me know any issues.