Setup is JH 1.4.1 + Jupyter Lab (different versions, see below). Rstudio Server is build as a module and loaded dynamically with LMod (if that does not make sense, a glimpse at this video is better than a ton of text: Dynamic libraries and applications for JupyterLab - YouTube). I’m using jupyter-server-proxy and jupyter-rsession-proxy (well, identical configuration but embedded into my code at the moment).
I had a previous version of JL based on Fedora 32 (Python 3.8.7) with older versions of packages, like email@example.com… I am now based on ubi8 (Python 3.8.6), with newer versions of most packages, like firstname.lastname@example.org.
Now… My new JupyterLab image is working perfectly standalone on my desktop. I can launch it, load RRtudio and launch it, it’s proxied perfectly through JL.
Pushing this on my OpenShift environment, I’m using JupyterHub to launch JL. With the older JL image it works perfectly. With the new one (so same JH, same RStudio module), RStudio is failing at launch with “status code 599”.
I know the difference may come from packages versions, or maybe some config. I tried to swap almost everything, proceeding by elimination, cannot find the cause.
So what I’m looking for now is a way to debug this properly, as I don’t have any specific log anywhere. I don’t see any error in JupyterHub, and in the JupyterLab pod I only see “[E 2021-09-01 21:21:15.439 SingleUserLabApp log:181] 599 POST /user/admin/rstudio/rpc/client_init (admin@::ffff:10.129.2.5) 781.82ms” in the console log.
How/where can I see if there is an error from rserver itself?
- Launcher configuration: s2i-lmod-notebook/packages/jupyterlmodlauncher/jupyterlmodlauncher at ecef383015302b03958a5a0c0125init.py#L33
- RStudio Server@1.3.959
Thanks for any help with this!