Azure ML Workspace - Can a Compute Instance be Shared Between Multiple Users?

Vishal Kumar 105 Reputation points
2026-06-16T04:53:57.0966667+00:00

Hi Team,

I have a question regarding compute access in Azure Machine Learning Workspace.

In our Azure ML Workspace, I created a Compute Instance for model training. Both I and my colleague can see the compute resource in the workspace, but only I can use it. My colleague is unable to start or connect to the compute because it was created under my account.

Similarly, when I create another Compute Instance and assign it to my colleague, they can use it, but I am no longer able to use that compute.

My questions are:

  1. Is it possible for multiple users within the same Azure ML Workspace to share and use the same Compute Instance?
  2. If not, what is the recommended approach for collaborative model development and training?
  3. Should we be using a Compute Cluster instead of Compute Instances for shared access scenarios?
  4. Are there any RBAC permissions or workspace settings that allow multiple users to access the same compute resource?

Any guidance on the best practice for sharing compute resources among team members would be appreciated.

Thanks.

Azure Machine Learning
Answer accepted by question author
Anshika Varshney 15,625 Reputation points Microsoft External Staff Moderator
2026-06-18T12:17:15.3966667+00:00

Hi @Vishal Kumar ,

Azure ML Compute Instances are designed for single-user access. Only the creator of the compute instance (or the user it was created on behalf of) can access and run notebooks, Jupyter Lab, VS Code, or RStudio on that instance.

So, while other users in the workspace may be able to see the compute instance, they won't be able to use it for interactive development.

Recommended approach for collaboration

  • Each team member should create and use their own Compute Instance.
  • Share notebooks, scripts, and datasets through the Azure ML workspace, shared storage, or a source control system such as Git.
  • For shared training workloads, consider using a Compute Cluster, which is designed for submitting jobs from multiple users.

Regarding RBAC

RBAC permissions can control who can create, start, stop, or delete compute resources, but they do not override the single-user access model of a Compute Instance. Even users with Contributor permissions cannot run notebooks on another user's Compute Instance.

Summary

  • Compute Instance = Single-user development environment
  • Each user should have their own Compute Instance
  • Shared files/notebooks can be stored in workspace storage
  • Compute Clusters are typically used for shared training and batch workloads

References

I hope this helps clarify the expected behavior. Do let me know if you have any further queries.

Thankyou!

Was this answer helpful?

1 person found this answer helpful.
0 comments
Answer accepted by question author

AI answer

2026-06-16T04:54:14.9666667+00:00
  1. A compute instance in Azure Machine Learning cannot be shared for interactive use between multiple users. It is a single-user, dedicated workstation. Only the creator (or the user it is explicitly assigned to) can access Jupyter/JupyterLab/RStudio and sign in to that compute instance. The instance has single-user sign-in and all actions use that user’s identity.
  2. For collaborative model development and training, the recommended pattern is:
    • Use individual compute instances as personal development environments for each data scientist.
    • Share code, notebooks, and data via the workspace (e.g., shared storage, repositories) rather than sharing the compute instance itself.
    • Use shared training compute targets (such as compute clusters) for running training and inference jobs that multiple users submit.
  3. Yes, a compute cluster is the recommended option for shared access scenarios. A compute cluster is a managed, shareable compute resource in the workspace that can scale up and down automatically and can be used as a training compute target by multiple users.
  4. Azure RBAC controls who can create, delete, start, stop, and restart compute instances and other compute resources in the workspace. All users with workspace Contributor or Owner roles can manage compute instances across the workspace. However, RBAC does not change the single-user nature of a compute instance’s interactive access: only the creator/assigned user can access Jupyter/JupyterLab/RStudio on that instance. Other users cannot “log into” or interactively use that same compute instance, even with RBAC permissions.

For shared compute among team members, use:

  • Compute instances for per-user development environments.
  • Compute clusters (or other shared compute targets) for multi-user training and inference workloads.

References:

AI-generated content may be incorrect. Read our transparency notes for more information.

Was this answer helpful?

1 person found this answer helpful.

0 additional answers

Sort by: Most helpful
Most helpful Newest Oldest

Your answer

Answers can be marked as 'Accepted' by the question author and 'Recommended' by moderators, which helps users know the answer solved the author's problem.