Appendix: DAG features

Click DAG next to a node to open its DAG view. The DAG interface has a central canvas that displays the node graph, a node list on the left for selecting nodes, and controls for features like node aggregation, upstream analysis, and downstream analysis at the top. You can perform the following operations in the DAG view. This topic uses an auto triggered instance as an example.

• Node aggregation

  If an auto triggered instance has a large number of nodes or dependency levels, you can aggregate nodes by dimensions such as node status, workspace, owner, or priority to view the count of nodes for a specific dimension. This helps you understand node distribution and manage execution more effectively. The following examples show the display when nodes are Not Aggregate and when they are aggregated by priority.

  Note

  You can aggregate nodes by status only in an instance's DAG. An instance can be an auto triggered instance, a data backfill instance, or a test instance.

  • When nodes are not aggregated, each node is displayed individually in the DAG. Each node shows its name and type, such as virtual node, ODPS SQL, or batch synchronization. Lines between nodes indicate dependencies. A message at the bottom indicates the current view contains 10 nodes across 8 dependency levels.

  • When nodes are aggregated by priority, the display is as follows. You can quickly see that the current node has 6 child nodes with a priority of 1. The aggregated DAG shows the dependency path, starting from fanyi_test_root (virtual node) at the top, passing through User Behavior Analysis Case Description (virtual node), table_ddl_ods_raw_log_d (ODPS SQL), two aggregated groups each containing two nodes, then to dws_user_info_all_di (ODPS SQL), ads_user_info_1d (ODPS SQL), and finally to User Behavior Analysis Result Query (ODPS SQL). Arrows between the nodes indicate scheduling dependencies. The left panel lists all task names, and you can select nodes by using checkboxes.

• Upstream analysis

  If a node has been pending for a long time, you can perform an upstream analysis. This reveals the blocking upstream nodes, allowing you to quickly locate and resolve runtime errors to improve execution efficiency.

  Note

  • You can use the upstream analysis feature only in the DAG of an instance. An instance can be an auto triggered instance, a data backfill instance, or a test instance.

  • You can perform an Upstream Analysis only on nodes that are not running. A maximum of six levels are displayed. If you want to view more levels, click Continue Analysis in the upper-left corner.

  • Upstream analysis helps you quickly find the ancestor nodes that the current node depends on but have not successfully run. These nodes might be in the Running, Failed, Pending (Time), Pending (Resource), or Frozen state. Ancestor nodes in these states block downstream execution.

  If an example node has not run for a long time, you can select the node and click Upstream Analysis in the upper-left corner. The analysis result shows that the blocking upstream node is the old_work_root node. With this result, you can quickly locate the blocking node and handle any related errors.

  Note

  Next steps:

  • After you locate the key instance, you can filter by an aggregation dimension to view the workspace or owner of the key instance that affects the data output of the current node. This improves operational efficiency.

  • After you locate the blocking instance, you can also right-click it and select Run Diagnostics to quickly analyze why it has not run successfully. For more information, see Intelligent diagnosis.

  In the upstream analysis results, old_work_root is a virtual node, and its downstream dependent node is the sd node of the ODPS SQL type.

• Downstream analysis

  By default, the DAG view displays three levels: the current node, its immediate ancestor nodes (one level up), and its immediate descendant nodes (one level down). When an auto triggered instance has a large number of nodes or levels, it can be difficult to see the full scope of impact. You can use the downstream analysis feature to filter by criteria such as node status, workspace, owner, or priority. This helps you count the number of nodes at different levels or the total number of nodes across all levels for a specific dimension.

  Note

  • Merged or layered view in an aggregated state: After you click Downstream Analysis, nodes are aggregated by owner by default, and the total number of nodes for that dimension across all levels is calculated.

  • Grouped view without aggregation: If there are more than 10 downstream nodes, they are displayed in groups by default. This view provides a more intuitive understanding of the nodes' business impact.

  • Ungrouped view: In the non-aggregated view, you can click Ungroup to expand the grouped nodes. This allows you to see the upstream and downstream relationships of all nodes.

  • The Layer view for a Downstream Analysis shows a maximum of six levels. To view more levels, click Continue Analysis in the upper-left corner.

  The following example shows the results of performing a Downstream Analysis on the tag node:

  • Merged view: Merges the analyzed downstream nodes based on the current aggregation dimension. In a non-aggregated state, downstream nodes are displayed in groups. In this example, the Merge method is used and nodes are aggregated by workspace. All downstream nodes are merged into a single layer, showing the number of nodes belonging to different workspaces. After you click the Merged button in the DAG view, the downstream analysis result for the tag node is displayed in merged mode. The seven downstream nodes are shown in aggregated groups, with each group containing a different number of descendant nodes. The status bar at the bottom indicates that 42 nodes and 5 levels are currently displayed.

  • Layered view: Displays the analyzed downstream nodes by dependency level based on the current aggregation dimension. In this example, the Layer method is used and nodes are aggregated by workspace, showing the number of nodes in different workspaces at different levels. After you select the DAG view and click the Layered button, the downstream nodes are arranged from top to bottom by level. Each level shows the number of nodes it contains, and the bottom of the page displays the total number of nodes and levels shown (for example, "42 nodes and 5 levels are displayed").

• DAG display styles

  You can click the function icons in the upper-right corner of the DAG view to adjust its display style. For example, you can use Toggle Full Screen View or Fit Screen.

  The following examples show the DAG for the downstream nodes of a node in Ungroup and Group views.

  • Ungroup DAG: This view helps you clearly understand the upstream and downstream relationships of all nodes. In ungrouped mode, each downstream node is displayed as an individual card arranged by dependency level, showing a total of 15 nodes and 2 levels. You can click the Snapshot icon in the right-side toolbar to save the current DAG as an image.

  • Group DAG: This view displays downstream nodes in an orderly manner, with five nodes per row, making it easy to quickly see the total number of downstream nodes.

  • Ungroup DAG: This view helps you clearly understand the upstream and downstream relationships of all nodes. In ungrouped mode, each downstream node is displayed as an individual card arranged by dependency level, showing a total of 15 nodes and 2 levels. You can click the Snapshot icon in the right-side toolbar to save the current DAG as an image.

  • Group DAG: This view displays downstream nodes in an orderly manner, with five nodes per row, making it easy to quickly see the total number of downstream nodes.