Backfill data for a node-阿里云帮助中心

Use natural language commands with code agents like Qoder and CodeX to backfill data for Dataphin scheduling nodes by business date. You can backfill historical data, correct output for specific periods, and track the status of data backfill workflows and instances. This article covers the skill's features, prerequisites, usage examples, and expected results.

Feature overview

With the Backfill Data for a Node skill, you can use natural language commands to backfill historical data for Dataphin scheduling nodes and their downstream dependencies by business date. Common use cases include initiating a data backfill for a specific node, backfilling data for a node and all its downstream nodes (massive mode), specifying the run times for hourly tasks, and tracking the run status of data backfill workflows and node instances for each business date.

Use cases:

Backfill historical data for a node by business date, or fix the output data for a specific period.
Backfill data for a node and all its downstream nodes to fix the historical output of an entire pipeline.
For hourly scheduled tasks, backfill data only for specific run times.
Track the execution status of workflows and node instances for each business date in a backfill run.

Note

Limitations: The backfill scope is calculated by business date, which defaults to T-1 (yesterday), not the scheduling day. A node ID maps to a single environment. For a project that has not been published to production, you must use the development environment to backfill data. Otherwise, Dataphin cannot find the node due to an environment mismatch. A backfill run is limited to the current project and cannot span multiple projects. After submitting a backfill request in massive mode, you must query the submission status to get the actual backfill ID, which is required to track workflows and instances.

Prerequisites

Have the target tenant ID and project ID.
Know the target node and its deployment environment (development or production).
Know the business date range to backfill.

Examples

You can describe your requirements to the agent by using natural language. Here are some examples:

Backfill yesterday’s data for a Shell task by business date.
Backfill data for a node and all of its downstream nodes.
Backfill historical data for a specific node from the 1st to the 7th of this month.
For an hourly task, backfill data only for the 08:45 and 20:45 run times.
Query the run status for each business date in a backfill run.
View the execution details of each node instance within a data backfill workflow.

Interaction steps:

The agent typically starts by confirming the target node, deployment environment (development or production), and the business date range to backfill (defaults to T-1, yesterday).
The agent typically locates the node by task type and name, then confirms its deployment environment to avoid a "node not found" error caused by an environment mismatch.
Before submitting the backfill request, the agent confirms the node, environment, business date range, whether to include all downstream nodes, and the concurrency limit.
After submitting a request in massive mode, the agent queries the submission status to get the actual backfill ID, then uses that ID to track the workflow and instance status for each business date.

Execution results

The operation returns the following results:

Response content: Submitting a data backfill request returns a submission ID. Tracking queries return the status (such as Initializing, Waiting, Running, Success, or Failed) and duration of the workflow for each business date. Instance queries return the status, run time, and duration of each node instance.
Completion status: The workflow and instance statuses are the definitive record. A Success status indicates the data backfill for that business date is complete.
Review and validation: Use the data backfill workflow overview and instance details to verify the execution status for each business date and node. If a run remains in the Waiting state for a long time, check the instance status to troubleshoot scheduling dependency issues or resource queuing.

Note

A "success" response from the command line only indicates that the system has accepted the request. Always rely on the actual results shown in the Dataphin console.

Input parameters

The following table lists parameters for initiating a data backfill. The agent automatically identifies the necessary information and prompts you for any missing details. You do not need to provide system-generated parameters.

Parameter	Required	Description	Example
Tenant ID	Conditionally required	Optional if specified in the credential configuration.	1234567890123456789
Project ID	Yes	The project to which the target node belongs.	—
Environment	No	The deployment environment, which can be development or production. The default is production. This must match the actual deployment environment of the node.	production
Node	Yes	The root node for the data backfill. You can specify multiple nodes.	—
Business date range	Yes	The start and end business dates (inclusive). Defaults to T-1 (yesterday).	20260901 to 20260907
Include all downstream nodes	No	Specifies whether to use massive mode. This mode automatically includes all downstream nodes of the root node.	No
Concurrency limit	No	The maximum number of concurrent instances for a single business date. Default: 1.	1
Run time range	No	For hourly tasks only. Specifies the run times to backfill.	08:45 to 20:45