When a branch node has mutually exclusive branches, exactly one branch runs and the other receives the status Branch Not Selected. If a downstream node depends directly on both branches, it inherits Branch Not Selected from the skipped branch and enters a dry-run state — the node is considered to have run, but its task is never executed. The same applies to all of that node's descendants.
A merge node solves this problem. It consolidates the run statuses of multiple upstream branch nodes into a single dependency point. Downstream nodes depend on the merge node instead of the individual branches, so a skipped branch no longer causes downstream nodes to be skipped.
Prerequisites
Before you begin, make sure you have:
A RAM user added to your workspace with the Develop or Workspace Administrator role. The Workspace Administrator role grants more permissions than required — assign it with caution. For details, see Add workspace members and assign roles to them.
A serverless resource group associated with your workspace. For details, see Use serverless resource groups.
A merge node already created. For details, see Create an auto triggered task.
Supported editions
Merge nodes are available in DataWorks Standard Edition and higher. For details, see DataWorks editions.
How it works
The following example illustrates the problem and solution.
Branch node C has two mutually exclusive branches, C1 and C2. Both write data to the same MaxCompute table. Downstream node B needs the output from that table.
| Setup | Behavior |
|---|---|
| Node B depends directly on C1 and C2 | One branch always gets Branch Not Selected, so B enters a dry-run state and never executes. |
| Merge node J merges C1 and C2; node B depends on J | B runs correctly after whichever branch completes. |
Step 1: Define the merge logic
On the node configuration tab, go to the merge logic definition section. Search for the branch nodes you want to merge by node output, node ID, or node name.
Click the
icon to add a node to the Merge Condition Settings section. Repeat for each branch node you want to include.In Merge Condition Settings, configure the merge logic condition and the run status condition for each branch node.
Merge logic conditions:
Condition Behavior AND The merge node is marked Successful only if all upstream branch nodes have finished running and each meets its specified run status. OR The merge node is marked Successful if all upstream nodes have finished running and at least one branch node meets its specified run status. Run statuses available for each branch node:
Status Meaning Successful The node ran successfully. Failed The node failed to run. Branch Not Run The node was not selected to run (dry-run state). It is considered to have run successfully, but its task was not executed. Applies only when the ancestor node is a branch node. In Result Settings, set the run status for the current node. The figure above shows an example configuration:
Branch 1 condition: Successful, Branch Not Run, or Failed — the condition is met as long as Branch 1 finishes running, regardless of outcome.
Branch 2 condition: Successful or Branch Not Run — Branch 2 must finish without failing.
Merge logic: AND.
Result: the merge node is marked Successful only if Branch 1 has finished running AND Branch 2 has finished running without failing.
You cannot select a run status for the merge node itself. You can only set the run status to Successful.
In the right-side pane, click Scheduling Properties to configure the scheduling settings. For details, see Configure scheduling properties.
Step 2: Deploy the node and perform O&M
Submit and publish the merge node to the production environment. For details, see Deploy a node or workflow.
After publishing, the task runs periodically based on your scheduling configuration. Go to Operation Center > Node O&M > Auto Triggered Node O&M > Auto Triggered Nodes to view and manage the task. For details, see Get started with Operation Center.