What causes WORKFLOW_DEPENDENCY_CYCLE?

Task A depends on Task B and Task B depends on Task A, directly or through an intermediate task. A copy-paste error when adding a new task inadvertently replicated a depends_on reference to an upstream task. A refactoring of task order introduced a reverse dependency that was not removed. Autogenerated bundle YAML from a script produced an incorrect depends_on list

How do I fix WORKFLOW_DEPENDENCY_CYCLE?

Step 1: Draw or export the task dependency graph from the Databricks Jobs UI (Runs tab, DAG view) to identify the cycle.. Step 2: Review the depends_on blocks in each task definition to find the task that closes the cycle.. Step 3: Remove or redirect the dependency that creates the cycle — if two tasks are mutually dependent, refactor them into a single task or introduce an intermediate task.. Step 4: If using DABs, run 'databricks bundle validate' after the fix to confirm the cycle is resolved.. Step 5: Save and run the job to confirm the DAG is now a valid directed acyclic graph.

High severityconfiguration

Power BI Refresh Error:
WORKFLOW_DEPENDENCY_CYCLE

What does this error mean?

A Databricks job or Declarative Automation Bundle definition contains a circular dependency between tasks, making it impossible to construct a valid execution order.

Common causes

1Task A depends on Task B and Task B depends on Task A, directly or through an intermediate task
2A copy-paste error when adding a new task inadvertently replicated a depends_on reference to an upstream task
3A refactoring of task order introduced a reverse dependency that was not removed
4Autogenerated bundle YAML from a script produced an incorrect depends_on list

How to fix it

1Step 1: Draw or export the task dependency graph from the Databricks Jobs UI (Runs tab, DAG view) to identify the cycle.
2Step 2: Review the depends_on blocks in each task definition to find the task that closes the cycle.
3Step 3: Remove or redirect the dependency that creates the cycle — if two tasks are mutually dependent, refactor them into a single task or introduce an intermediate task.
4Step 4: If using DABs, run 'databricks bundle validate' after the fix to confirm the cycle is resolved.
5Step 5: Save and run the job to confirm the DAG is now a valid directed acyclic graph.

Frequently asked questions

Can Databricks detect the cycle automatically before the job runs?

Yes — the Jobs API and the bundle validate command both reject job definitions with circular dependencies at submit time, so the error appears before any task executes.

How should I handle tasks that need to share data bidirectionally?

Use a shared Delta table or a Unity Catalog volume as the communication layer rather than task dependencies. Write intermediate results to the shared location and have each task read from it independently, breaking the circular reference.