Reputation: 53
When using a GCP AI Platform Jupyter Lab I am experiencing a unresponsive Jupyter Notebook.
The jupyter notebook is accessed over https (using the Open Jupyterlab button in GCP AI Platform).
The notebook says "Kernel Connecting" for several minutes before saying "Kernel Disconnected" in the status window. When inspecting the browser error messages, it is reporting the below error message:
WebSocket connection to 'wss://475211ea312377e9-dot-us-east1.notebooks.googleusercontent.com/api/kernels/e77c9731-fc84-48d0-89ae-dfba685c37c1/channels?session_id=a6e7c6a9-ac86-4723-b435-74ca34036ded' failed: _createSocket @ default.js:73
I am using the basic Python 3 image with no special settings.
I am accessing the jupyter notebook from a Windows 10 Pro Corporate computer behind a corporate firewall / proxy.
When trying to access the jupyter notebook on an android phone, the jupyter notebook works fine, but when using company laptop, it is not working. I have tried modifying several firewall settings, but still not able to connect.
Has anyone else experienced a similar issue? Any ideas on how I can resolve the issue and connect via Windows corporate laptop?
Upvotes: 1
Views: 2609
Reputation: 536
There is a simple Remote Utility to connect to the Jupyter Notebook. GCP recommend to use SSH to connect Kernel from remote.
Upvotes: 1
Reputation: 10058
You should take a look at WebSockets connections, as your test implied the connections are blocked by your Firewall. I would recommend to open a case with the Firewall vendor, check Websocket settings. If WS is not supported, I would use SSH to connect to it.
https://cloud.google.com/vertex-ai/docs/workbench/user-managed/ssh-access
Upvotes: 1