Problem with real-time collaboration extension on Jupyterhub

JupyterHub
1

Hi there, I’ve installed the real-time collaboration extension on Jupyterhub. The basic workflow seems to work correctly, but I’m experiencing 3 issues:

  1. When I open a Jupyter notebook from before the installation of RTC, it does not show any content (see picture), (for a newly created notebook, RTC works flawlessly).
  2. When I click on the ‘open as notebook’ icon (also in the picture with a red circle), it tries to open the URL https://100.74.57.128/user/erco/tree/RTC:cbdd/RandomForest/Pipeline/cbdd.ipynb, so with a semicolon, which does not work. If I manually replace the semicolon with its escape sequence (https://100.74.57.128/user/erco/tree/RTC%3Acbdd/RandomForest/Pipeline/cbdd.ipynb), it still does not work. If I completely remove ‘RTC%3A’ from the URL, the URL opens, but I seem to have lost real-time collaboration capability. The fact that the ‘open as notebook’ does not function correctly, is confusing to users.
    Update: there’s already a number of problem reports written on this issue (RTC breaks notebook links · Issue #202 · jupyterlab/jupyter-collaboration · GitHub, Nbconvert jumps to the wrong link in collaboration mode · Issue #151 · jupyterlab/jupyter-collaboration · GitHub, Unable to Use Search and Replace with Real Time Collaboration because of No Support for Custom Drives · Issue #107 · jupyterlab-contrib/search-replace · GitHub)
  3. Sometimes there are strange sequences of characters at the prompt of an opened x-terminal. Example: 2RR0;276;0c10;rgb:0000/0000/000011;rgb:ffff/ffff/ffff12;2$y. They are not always there. I don’t understand where they are coming from. It doesn’t affect functionality, but is confusing to users.

cbdd_problem
cbdd_problem1918×535 35.7 KB

versions of packages:

versions
versions743×412 7.56 KB

in the log you can see that I’m opening the Jupyter notebook and you can also see that RTC performs an autosave:

log1
log11623×862 104 KB

1 Like
2

At first, it seemed that removing ‘RTC%3A’ from the URL allowed users to continue on working, but several times the jupyterhub-singleuser process of a user went to 100% CPU, with stopping the server process as only solution.

So I’ve removed the RTC extension, which is a pity, as groups of students are using this server for their projects and they would really benefit from the RTC capability.

3

If you can create a minimal reproducible example it’d be worth opening an issue on

1 Like