UC Model Deployment across data bricks instances

srkam — Wed, 10 Dec 2025 17:04:13 GMT

Hello, We have multiple data bricks instances each represents an environment dev,qa,rel,prod etc.. We developed a model in the dev workspace and registered in the UC model registry using mlflow. Now, we are trying to find a best way to deploy this registered model into the target environments. We want to avoid rerun of the training pipeline in the target env, instead promote/copy/transition the registered model, its version and experiments into the target env.

Can you please help us on how to achieve this?

Thanks

Re: UC Model Deployment across data bricks instances

iyashk-DB — Thu, 11 Dec 2025 03:54:10 GMT

You can use UC's centralized model registry and MLflow’s copy APIs.

If all target workspaces attach to the same Unity Catalog metastore, reference and promote models via their 3‑level UC names; use MLflow’s copy_model_version to “copy” the exact artifacts from dev to qa/rel/prod, and manage deployment with aliases like Champion/Shadow. This avoids retraining and keeps one source of truth.
Ref Doc - https://docs.databricks.com/aws/en/machine-learning/manage-model-lifecycle

If environments run on different, isolated metastores/workspaces, use the community mlflow-export-import tooling to migrate registered models, versions, and experiments/runs between workspaces. This is the recommended way to copy MLflow objects (models, runs, experiments) across workspaces when UC sharing isn’t possible.
Ref Doc - https://github.com/mlflow/mlflow-export-import

topic Re: UC Model Deployment across data bricks instances in Machine Learning

UC Model Deployment across data bricks instances

Re: UC Model Deployment across data bricks instances