Hello! I am working in a deployment of Jupyterhub on GCP (using the Pangeo configuration). Occasionally (last Friday, and again today) Jupyterlab seems very slow to load up. Everything up the point at which the JLab interface should show up in the browser seems to work as normal – server startup runs smoothly, but when I am redirected to http://learning2learn.pangeo.io/user/arokem/lab?, the background goes white and remains white for up to an hour with no interface appearing. Any ideas what might be going on?
More information here: this is definitely a JupyterLab issue: when I go to http://learning2learn.pangeo.io/user/arokem/tree instead, I can see all my notebooks and run things there. Also, I checked the pod logs with kubectl and there’s nothing out of the ordinary there. To be on the safe side, I disabled lab extensions running on this cluster. But still no joy.
Do you have any tips for how to reproduce this behavior @arokem? I’m not sure what the issue might be, but things seemed fine when I logged in.
Perhaps there is some corruption in the restore state that is causing the issue. The next time this happens, can you try to navigate to
/lab?reset to see if that fixes it?
Thanks for helping. Very hard to reproduce – it’s intermittent on my end as well. But I will definitely try the
lab?reset url. What does that do?
Basically, JupyterLab keeps a JSON file on disk that stores state-related stuff (layout, principally, but some other stuff as well). Sometimes that can get messed up for whatever reason (bug, misbehaving extension, etc), and can interfere with loading of the application. Adding the reset parameter throws away the JSON file and starts fresh.
Thanks! That’s super useful to know. I’ll wait until it happens again and then try this and report back whether it solves the issue.
Just here to say that indeed, it worked in at least one case. Thanks again!