Jupyter Notebook won't load in ML Studio

Question

Jupyter Notebook won't load in ML Studio

Katelan 0

Hi all!

I'm having troubles with loading any of my .ipynb notebooks in ML Studio. I can navigate through other files like CSVs, but when I try to open a jupyter notebook, the entire page becomes unresponsive and completely frozen. When I try to refresh my page, it takes me to a connection error.

Has anyone encountered a similar issue?

Manish Deshpande 7,520 Reputation points Microsoft External Staff Moderator

2026-07-03T18:15:26.8+00:00
Hi @Katelan ,

Thanks for the detailed report a notebook that freezes the whole page and then throws a connection error on refresh (while everything else in the workspace loads fine) usually points to something on the notebook/kernel side rather than the workspace itself. Here's what I'd try, roughly in order of least to most involved:

Rule out the browser first

Open the notebook in a private/incognito window, or a different browser. A stale cache or extension is a very common cause of a tab freezing specifically on .ipynb files. If that works, clearing your cache/cookies for ml.azure.com should fix it for good.

Check the compute instance state

Make sure your compute instance shows green (Running) on the Compute page before opening the notebook — an instance that's still starting or "Setting Up" can make the notebook editor hang indefinitely. If it's not green, restart it from Compute > ... > Restart:

https://learn.microsoft.com/en-us/azure/machine-learning/how-to-manage-compute-instance

Try opening the same file in classic Jupyter or JupyterLab

Once compute is running, use the notebook toolbar to launch the file directly in Jupyter or JupyterLab instead of the native studio editor. If it opens fine there, the issue is isolated to the studio notebook UI (often triggered by a notebook with very large cell outputs), and you can keep working from Jupyter/JupyterLab while we chase the root cause:

https://learn.microsoft.com/en-us/azure/machine-learning/how-to-run-jupyter-notebooks#edit-a-notebook

Check for a kernel crash

If the kernel crashed and silently restarted, that would explain the freeze + connection error combo. Open a terminal on the compute instance and run:

`sudo journalctl -u jupyter`

That shows the Jupyter service log. If it points to an out-of-memory kill, moving to a compute instance with more memory usually resolves it. Terminal access steps:

https://learn.microsoft.com/en-us/azure/machine-learning/how-to-access-terminal

Sign out and back in

An expired auth token mid-session can leave a notebook stuck exactly like this. Sign out of ML Studio, sign back in, then reopen the notebook and restart the kernel.

Network/firewall (only if you're on a VPN or corporate network)

The notebook editor depends on WebSocket connections to *.instances.azureml.net and *.instances.azureml.ms. If a corporate proxy blocks WebSockets, you'll see this exact symptom — page hangs, then connection error on refresh. Worth checking with your network team against the required endpoints:

https://learn.microsoft.com/en-us/azure/machine-learning/how-to-access-azureml-behind-firewall

For reference, all of the above (plus a couple of other edge cases) are covered in the official troubleshooting section here:

https://learn.microsoft.com/en-us/azure/machine-learning/how-to-run-jupyter-notebooks#troubleshooting

If none of this clears it up, it'd help to know:

Whether this happens on every notebook or just this one

Any error in the browser console (F12 → Console tab) at the moment it freezes

Thanks,
Manish.

1 answer

Your answer

Answer 1

The behavior described is consistent with notebook connectivity issues in Azure Machine Learning studio that rely on WebSocket connections.

Use these checks and fixes:

Verify WebSocket connectivity
- Notebook execution and interaction require WebSocket communication.
- Ensure the network or firewall allows WebSocket connections to:
  - *.instances.azureml.net
  - *.instances.azureml.ms
- If behind a corporate firewall/proxy, ask the network administrator to confirm outbound WebSocket traffic to those hosts is not blocked.
Check private endpoint / VNet setup (if used)
- If the workspace uses a private endpoint and the compute instance is deployed in a virtual network, it can only be accessed from within that VNet.
- If using custom DNS or a hosts file, ensure there is an entry for:
  - <instance-name>.<region>.instances.azureml.ms pointing to the private IP of the workspace private endpoint.
Sign out and back in (expired token)
- If the connection error is related to an expired token, sign out of Azure Machine Learning studio, sign back in, and then reopen the notebook.
Check compute and kernel status
- In the notebook UI, verify the compute instance status indicator:
  - Green: running
  - Red: failed
  - Black: stopped
  - Light blue: creating/starting/restarting/setting up
  - Gray: deleting/stopping
- Also check the kernel status indicator:
  - Green: connected/idle/busy
  - Gray: not connected
- If compute or kernel is not running/connected, start the compute instance and wait for the kernel to connect before interacting with the notebook.
Try another network or machine
- If possible, open the same workspace and notebook from a different network (e.g., outside corporate network) or a different workstation.
- If it works elsewhere, this strongly indicates a local network/firewall or browser environment issue blocking WebSockets.
Check Jupyter logs if kernel crashes
- If the kernel is crashing and causing the page to hang, use:
  - sudo journalctl -u jupyter on the compute instance to inspect Jupyter logs.
- If kernel issues persist, consider using a compute instance with more memory.

If after these steps notebooks still freeze while other files (like CSVs) open fine, focus on WebSocket/network restrictions and any private endpoint/DNS configuration, as those are the primary causes of notebook-only loading issues.

References: