The issue with the 'Repair DAG' and 'Repair All DAGs' options being disabled in the Airflow UI when using the Databricks Workflow Operator is a known UI-specific problem that does not affect backend execution or the actual repair functionality. While tasks can be successfully repaired through workflow execution, the visual controls in the UI appear incorrectly disabled.
Root Cause and Current State
-
The backend repair functionalities, which rely on Airflow's integration with the Databricks API, continue to operate as expected, allowing DAGs and individual tasks to be repaired regardless of the UI state.
-
The cause of the visual issue appears to relate specifically to the way extra operator links (such as for repair actions) are surfaced and enabled in the Airflow UI for tasks that use custom Databricks integration classes, especially in newer provider versions or complex environments.
-
This issue is not unique to your setup and has surfaced for other users, suggesting a bug or missing UI state synchronization for these buttons when Databricks compute or operators are involved.
Possible Remediation and Best Practices
-
Ensure you are running the latest version of both Airflow and the apache-airflow-providers-databricks package, as several bugs related to Databricks repair and UI actions have been addressed in recent releases.
-
Review whether your Databricks connection and job/task parameters are correctly surfaced to the UI—missing run identifiers, connection IDs, or improper task states may lead to UI links being disabled.
-
If you need immediate access to these options, you can continue to trigger repairs via direct API calls, backend scripts, or through the Databricks UI, as the backend is not impacted.
Additional Notes
-
Monitor the Airflow and Databricks provider changelogs for further fixes to UI-related issues. Specifically, bugs related to "repair buttons not overriding template parameters" and UI state have been patched as recently as late 2025.
-
If running Airflow 2.x or an older provider, consider upgrading because some operator link behaviors (such as repair all failed tasks) are handled differently in Airflow 3.x and in new Databricks provider releases.
In summary, the disabled repair options in the UI are a frontend problem impacting user interaction but do not prevent actual job repair through Airflow or Databricks API calls. Keeping your environment updated is the best path toward a fix, but repairs can proceed through backend paths until an official UI update resolvesolves the issue
