rapidsai / jupyterlab-nvdashboard

A JupyterLab extension for displaying dashboards of GPU usage.
BSD 3-Clause "New" or "Revised" License
598 stars 78 forks source link

Correctly categorizing the card in the Launcher #98

Closed vidosits closed 3 years ago

vidosits commented 3 years ago

Hey, would it be possible to categorize the extension's launch card correctly so that it doesn't show up in Notebooks? image According to JupyterLab docs it should be possible:

jacobtomlinson commented 3 years ago

Strange I don't see this.

image

Could you share your Jupyter Lab version?

vidosits commented 3 years ago

Hey Jacob!

Jupyter's version is v3.0.15. nvdashboard's version is 0.6.0a2105117 as installed with pip install --pre jupyterlab-nvdashboard

image

jacobtomlinson commented 3 years ago

Interesting. I'm on 3.0.11 so it must've snuck in.

We don't actually add anything to the launcher explicitly so it must be discovering it somehow.

vidosits commented 3 years ago

Yeah, very strange. In the docs they seem to explicitly require the usage of ILauncher, but I haven't been able to find a reference to that token in this repo.

JupyterLab docs JupyterLab API docs

I also don't see any mention of Card or Launcher in the changelogs since 3.0.11.

Just to be sure, I also tested with an empty environment with Docker:

FROM continuumio/miniconda3
RUN apt update
RUN apt install -y gcc
RUN pip install --pre jupyterlab_nvdashboard
ENTRYPOINT ["jupyter", "lab", "--allow-root", "--ip=0.0.0.0"]

Built and ran with: docker build -t test:latest . && docker run --rm -p 8888:8888 test:latest

vidosits commented 3 years ago

Doesn't seem to be related to Jupyter's minor version, I can also reproduce it on 3.0.11: image

FROM python:3.7
RUN apt update
RUN apt install -y gcc
RUN pip install jupyterlab==3.0.11
RUN pip install --pre jupyterlab_nvdashboard
ENTRYPOINT ["jupyter", "lab", "--allow-root", "--ip=0.0.0.0"]

Built and ran with: docker build -t test:latest . && docker run --rm -p 8888:8888 test:latest

jacobtomlinson commented 3 years ago

How strange.

@ian-r-rose I know you don't work on this project but do you have any idea what is going on here? Are there any docs you could point me to?

vidosits commented 3 years ago

@jacobtomlinson Have you been able to reproduce it on your end (maybe w/ Docker?). It's really strange that JupyterLab v3.0.11 gives a different setup to you than for me, although I'm not sure if you've checked with v0.6.0 of this project.

ian-r-rose commented 3 years ago

This is an effect of the jupyterlab-server-proxy labextension (note the "lab", it's independent of the `jupyter-server-proxy server extension). Some of the maintainers there like the idea of putting proxied applications on top, and hack it into the notebooks category. See this issue, started by yours truly.

I usually avoid installing the jupyterlab-server-proxy package for this reason. It would be best if somebody could figure out an upstream fix that satisfies all parties, though I haven't weighed in recently, sorry to say.