Workspace URL prefix generation failing, causes Che-in-Che

[gchang]

No description provided.

[gchang]

The issue is this code block here: https://github.com/MAAP-Project/maap-workspaces/blob/e2302d9c5b9675a35772418c5cac6b74a570936d/jupyterlab3/entrypoint.sh#L3-L62

At the high level, what it's doing is trying to get the URL that will serve up the jupyter interface. It's in the form of https://blahblahblah/serverdbq97vva-ws-jupyter/server-3100. It's calling the Kubernetes API from inside the pod to get the strings serverdbq97vva and the port 3100. The problem seems to be on occasion -- maybe it's many workspaces running, or having concurrent workspaces starting up -- the Kubernetes API fails to return a value. That's why there's a retry loop in there, current set to 20 retries.

However, there isn't a sleep there, so if it fails, it cycles through the 20 iterations really quickly and then proceeds without the correct URL prefix. As such, then it loads the root page, which is Che itself.

I don't believe there's a way to "refresh" the Kubernetes API to ensure that it returns a value, so the only way is to wait longer. The proposed fix will have a sleep in there to give it some delay as well as increase the iterations.