A data engineering team manages multiple iterations of a machine learning model in Google Cloud's Model Registry. They are cleaning up obsolete assets and need to delete an older model version that is currently designated with the default alias. What must the team do in Model Registry to successfully delete this model version?

Assign the default alias to another model version prior to or during the deletion confirmation process.: ### **What Model Registry Versioning and Aliases Are** In Google Cloud's **Model Registry**, models are managed as versioned assets under a unified model resource. Each version represents a specific iteration of trained model artifacts, evaluation metrics, and metadata. To simplify model tracking, promotion, and downstream deployment, **version aliases** (such as `default` or custom labels like `production` and `staging`) point to specific numeric versions. The `default` alias identifies the primary version of the model used whenever a specific version ID is not explicitly provided in API calls or deployment pipelines. ### **How It Meets the Requirements** - **Protects Active Workflows (✓):** The Model Registry enforces safety guardrails preventing the direct deletion of a version holding the active `default` alias, ensuring downstream processes referencing the default model do not encounter missing resources. - **Enables Seamless Reassignment (✓):** When initiating deletion of the default version through the Google Cloud console or API/SDK, the user is required to select another existing model version to inherit the `default` alias, ensuring continuous model governance and integrity. - **Safe Asset Lifecycle Management (✓):** Once the default alias is safely re-pointed to another designated version, the obsolete version and its associated evaluation metrics can be safely purged. ### **Key Advantages and Governance Best Practices** - **Lifecycle Identifiers:** Aliases abstract mutable model states from immutable version numbers, allowing seamless promotion without code changes. - **Accidental Deletion Prevention:** Built-in safeguards protect critical operational versions from unintended deletion. - **Clean Version Lineage:** Deleting obsolete iterations keeps the registry organized while retaining required evaluation histories for active models. - **Unified Metadata Tracking:** Ensures that active models always retain proper provenance, schemas, and performance evaluations. ### **Why This Is Optimal** Reassigning the `default` alias ensures zero operational downtime for client services querying the model without version qualifiers. It is the established governance procedure in Google Cloud Model Registry to maintain asset consistency.