Azure Machine Learning
An Azure machine learning service for building and deploying models.
3,332 questions
Issue related with Compute Attach in Azure Machine Learning Studio. SSH authentication fail with the VM.
![](data:image/svg+xml, %3Csvg xmlns='http://www.w3.org/2000/svg' height='64' class='font-weight-bold' style='font: 600 30.11764705882353px "SegoeUI", Arial' width='64'%3E%3Ccircle fill='hsl(12.8, 46%, 11%)' cx='32' cy='32' r='32' /%3E%3Ctext x='50%25' y='55%25' dominant-baseline='middle' text-anchor='middle' fill='%23FFF' %3EAN%3C/text%3E%3C/svg%3E)
Aakash Nakarmi
0
Reputation points
I am not able to attach my VM(For Compute) to my Azure Machine Learning Studio. It is showing SSH authentication error.
However, I am able to access my VM through my local machine via SSH only.
Regarding my VM,
SSH configuration is all set, SSH connection is public and anyone can access it via port 22.
VM Series is STANDARD_D16LDS_V5. However, I have tried other VM series that support the compute for ML studio but still same issue of SSH.
Regarding my ML studio resource,
it is also not restricted to outbound network and it has contributor role so it can access other resources as well.
Azure Machine Learning
Azure Virtual Machines
Azure Virtual Machines
An Azure service that is used to provision Windows and Linux virtual machines.
9,013 questions
{count} votes
-
romungi-MSFT 48,906 Reputation points Microsoft Employee Moderator2024-12-05T15:25:15.0433333+00:00 @Aakash Nakarmi I believe you are using a compute outside Azure ML and it is running ubuntu. Since only ubuntu is supported for Azure ML attached compute.
I also believe the resource id is in correct format with the correct username used which has the SSH key or password setup. The error indicates that it may be an issue with authentication, have you used an SSH key authentication? If yes, is the private key correctly pasted from your local system where it works for you directly?
-
Aakash Nakarmi 0 Reputation points
2024-12-06T06:20:49.6533333+00:00 Hi @romungi-MSFT , Thank you for replying to my issue.
Yes I am using Ubuntu VM 22 version image.
Yes the resource ID is correct. It is in the format /subscriptions/<subcription-d>/resourceGroups/<resource-group>/providers/Microsoft.Compute/virtualMachines/<vm-name>
Yes, I am using the same SSH key with .pem extension that works when I am connecting to the VM via SSH from my local system.
-
romungi-MSFT 48,906 Reputation points Microsoft Employee Moderator
2024-12-06T11:15:59.9433333+00:00 I see that there could be an issue here, I see similar behavior. I have however tried to use password-based authentication and it works in that case. See my screenshot below where the successful attach is using password-based authentication. Please use this workaround until I hear back from the Azure ML team. Thanks!!
-
Aakash Nakarmi 0 Reputation points
2024-12-06T12:59:53.6833333+00:00 Hey @romungi-MSFT ,
I know that I can use the password method for authentication but the requirement is to use SSH authentication.Please reply to this thread if you get to know how to resolve this issue. I will be looking forward. Thanks!
-
Ameer Mohamed Aslam Sujah 0 Reputation points2025-04-11T15:04:08.5766667+00:00 Hello Aakash Nakarmi,
Thanks for the details and the screenshot. Let’s Troubleshoot Step by Step:
- Confirm the SSH username:
In Azure ML Studio, when attaching the compute, ensure the SSH username matches exactly the one set up on the VM.
- Use SSH public key, not password:
Azure ML requires SSH public key authentication (not password-based) when attaching a compute target manually.
- Make sure the SSH key has correct permissions on the VM.
- VM must allow Azure ML outbound IP (network restrictions):
Even though you said it's public, double-check that NSG rules allow SSH (port 22) from Azure Machine Learning IPs. You could temporarily allow all inbound from your subscription, or open port 22 to all IPs just for testing.
- Check VM host OS and compatibility:
While this usually isn't an issue, Azure ML Compute expects a Linux VM (Ubuntu preferred) and may not properly attach to Windows or custom hardened VMs.
- Use CLI to attach VM (optional)
- Last resort: Enable Azure ML to create its own VM.
If the VM is purely meant for ML compute, consider allowing Azure ML to create the VM itself — this avoids config mismatches.
Sign in to comment
Sign in to answer