DAB does not update MLflow Experiment tags on redeploy

[ergunbaris]

Describe the issue

Databricks Asset Bundles does not update MLflow Experiment tags during databricks bundle deploy. The mlflow.workload_creation_policy_id tag is applied only when the Experiment is created; changing or adding it in databricks.yml is ignored for an existing Experiment.

Configuration

resources:
  experiments:
    app_trace_experiment:
      name: /Workspace/Shared/app-trace-experiment
      tags:
        - key: mlflow.workload_creation_policy_id
          value: "<POLICY_ID>"

Steps to reproduce the behavior

1. Deploy the bundle without the mlflow.workload_creation_policy_id tag.
2. Add the tag to databricks.yml.
3. Run databricks bundle deploy -t <target>.
4. Inspect the existing Experiment settings or tags.
5. The new tag is not applied.

Expected Behavior

Experiment tags defined in databricks.yml are reconciled during bundle updates, including adding or changing mlflow.workload_creation_policy_id.

Actual Behavior

The tag is ignored when the Experiment already exists. Deleting and recreating the Experiment is required for the tag to be applied, which is destructive and unsuitable for Experiments containing existing runs or traces.

OS and CLI version

• CLI version: v1.15.0
• OS: MacOS

Is this a regression?

• Tested with v1.9.0 and v1.15.0

Debug Logs

I need to recreate the issue for sharing related logs. I hope you don't need these logs Very busy with work.