Rerun task instances
You can use natural language commands to direct code agents, such as Qoder and CodeX, to rerun Dataphin scheduling task instances. This feature allows you to rerun a single instance, multiple instances in a batch, or an entire downstream pipeline starting from a root instance. This topic describes the feature's capabilities, prerequisites, usage examples, and lists input parameters in the appendix.
Overview
The rerun task instance feature lets you use natural language commands to rerun Dataphin scheduling task instances. Common use cases include locating and rerunning single or multiple instances based on the task name and business date. You can also rerun a root instance and all its downstream pipeline to repair an entire data pipeline. After the rerun, you can check the latest status and run logs for the instances.
Use cases:
-
Manually rerunning a periodic instance that has failed, timed out during scheduling, or been suspended.
-
Rerunning multiple failed instances for a specific business date in a batch.
-
Rerunning an entire downstream pipeline starting from a root instance to repair data.
-
Rerunning only the failed instances within a downstream pipeline.
Rerunning a task does not guarantee a fix. The rerun will fail if the underlying problem lies in the task configuration or source data. In such cases, you must diagnose the root cause by checking the run log. When you rerun a single instance, its status changes. In contrast, rerunning a downstream pipeline adds new run records to the original instances instead of creating new ones, which requires a different verification method. If multiple instances match your request, the agent prompts you to select a single target and does not rerun all of them automatically.
Prerequisites
-
You have the target tenant ID and project ID.
-
You know the target task name and business date. The default business date is the previous day (T-1).
-
The target instance already exists and has a status such as Failed, Scheduling Timeout, or Suspended.
-
When rerunning a downstream pipeline, you must specify the root instance and the scope of the downstream tasks to include.
Usage examples
You can describe your requirements to the agent in natural language. The following are example commands you can use directly:
-
Rerun the failed instance of a task from yesterday's business date.
-
Rerun multiple failed instances for a specific business date in a batch.
-
Rerun a root instance and all its downstream pipeline.
-
Rerun only the failed instances in the downstream pipeline.
-
Check the latest status of the instance after the rerun.
-
If the rerun fails again, retrieve the run log to investigate the cause.
Interaction steps:
-
The agent typically starts by asking you to confirm the target project, task name, business date (defaults to the previous day, T-1), and environment (development or production).
-
The agent then locates the target instance. If tasks with the same name exist across multiple projects, the agent lists all matching instances and asks you to select a single target.
-
Before executing the rerun, the agent confirms details such as the instance, task name, project, environment, and business date. When rerunning a downstream pipeline, it also confirms the root instance, the scope of downstream tasks, and whether to include the root instance itself.
-
After rerunning a single instance, the agent checks its status to confirm if it has transitioned from Failed to Running or Succeeded. If it fails again, the agent retrieves the log to help you find the root cause.
-
Rerunning a downstream pipeline does not create a new instance. Instead, the agent checks the run log of the original instance to verify that a new run record has been added and completed successfully.
Execution results
After the operation is complete, you can expect the following outcomes:
-
Response content: A rerun command returns whether the operation was successfully initiated. A status query returns the latest status, start and end times, and duration. A log query returns all run records and their log content.
-
Completion status: A successful command initiation does not mean the instance has completed successfully. For a single instance rerun, success is confirmed when the instance status changes from Failed to Succeeded. For a downstream pipeline rerun, success is confirmed when each instance has a new, successful run record.
-
Verification: You can verify the result by checking the instance status and run log. If the rerun fails, use the log to troubleshoot common issues, such as data source permissions, missing destination tables or fields, or resource queuing.
Always check the status in the Dataphin UI for the final result.
Appendix: Input parameters
The following table lists parameters used for rerunning task instances. This is for reference only. The agent automatically identifies the required information from your command and will prompt you if any details are missing. You do not need to provide parameters that are not required because the system generates them or applies default values.
|
Parameter |
Required |
Description |
Example |
|
tenant ID |
Conditional |
Can be omitted if specified in your credential configuration. |
1234567890123456789 |
|
project ID |
Yes |
The project to which the instance belongs. |
— |
|
environment |
No |
The environment of the task. Valid values are development or production. The default is production. |
production |
|
task name |
Yes |
The name of the task to rerun. |
— |
|
business date |
No |
The business date of the instance. The default is the previous day (T-1). |
20260907 |
|
root instance |
Required when rerunning a downstream pipeline. |
The instance from which to start the downstream pipeline rerun. |
— |
|
downstream scope |
No |
The set of downstream instances to include in the rerun. For example, "All Instances" or "All Failed Instances". |
All Instances |