How do we know if an azureml image is being used?

Hi Guilherme Matheus,

Azure Machine learning is tightly coupled service. Deleting the resources created along AML service will make the AML service corrupt. You can verify AML bound dependent services will appear in overview page.please find below.

 image (2).png

I am not asking about deleting resources that are coupled. I am asking about the images that AML created in the Container Registry, because some of those images are being flagged by SFI. How do I know which image I can delete?

Hi Guilherme Matheus,

Thanks for clarifying I understand you’re specifically referring to container images stored in your Azure Container Registry (ACR) that were created by Azure Machine Learning (AML), and you want to identify which ones are actively in use before deleting the flagged ones.

Here’s how you can safely determine whether an image is still being used by AML:

1. Check AML Environments Linked to Images

Azure Machine Learning associates each job or endpoint with an Environment, which in turn references an image in your Container Registry. You can list all environments in your workspace using either:

az ml environment list --resource-group <rg-name> --workspace-name <workspace-name>

or with the Python SDK:

from azure.ai.ml import MLClient
ml_client = MLClient.from_config()
envs = ml_client.environments.list()
for env in envs:
    print(env.name, env.image)

This helps you identify which ACR images are currently referenced by environments. Any image not linked to an active environment is generally safe to delete.

2. Check Active Endpoints, Models, and Pipelines

Navigate to your AML workspace → Endpoints, Models, and Pipelines sections.

• Endpoints use images tied to environments for serving models.

Pipelines and jobs also reference these environments during execution. If an image is tied to a deployed endpoint or an active scheduled pipeline, do not delete it.

3. Review Job History and Recent Runs

In the Jobs tab, you can review recent runs and check which environments (and images) were used. If an image hasn’t been used in recent runs or is linked to environments no longer deployed, it’s likely safe to remove.

4. Validate Before Deletion

Before removing any image, double-check:

No running or queued jobs depend on it.

The environment referencing it is not used in a registered endpoint.

You can then clean up unused images programmatically using ACR commands such as:

az acr repository delete --name <acr-name> --image <image-name>:<tag>

Avoid deleting system-managed base images (used by Azure ML for built-in environments) removing these may cause job failures when creating new runs or environments.

Please refer this Azure Machine Learning Documentation.

I Hope this helps. Do let me know if you have any further queries.

Thank you!