Data Asset Governance provides a comprehensive solution for gracefully undeploying obsolete tasks and tables. This feature uses risk assessment, notifications, and process management to help you overcome challenges like identifying undeployment risks, handling active downstream dependencies, and understanding potential impacts. It also offers batch undeployment to improve efficiency.
Background
Unlike direct undeployment, graceful undeployment includes impact analysis and process management. You can also notify stakeholders at each stage.
Actions | Undeployment Method | Comparison | Description |
Direct undeployment | Find the task on the O&M page, Frozen the task, and then undeploy it directly. | This method does not show impacts or risks, increasing the chance of error. | |
Graceful undeployment | You can select objects, assess their impact and risks, and add them either manually or automatically. | The system provides an impact assessment for objects selected by node ID, data lineage, or table. | |
Direct undeployment | None | You cannot notify relevant personnel. | |
Graceful undeployment | You can notify designated personnel at each stage of the process by using System Prompt, Mail, DingTalk Chatbot, and Webhook. | It enables a multi-stage process with notifications for relevant personnel at each step. | |
Direct undeployment | None | The object is undeployed immediately, and the action is irreversible. | |
Graceful undeployment |
| You can choose different execution modes for different scenarios, providing a window to cancel the undeployment if needed. | |
Direct undeployment | None | This immediate action lacks a buffer period, increasing the risk of accidental undeployment and business failures. | |
Graceful undeployment | You can view the details of created undeployment plans and perform operations such as viewing execution details, canceling, and rolling back for each subtask. | It provides a buffer for undeployment, allowing you to cancel or roll back the process if an error occurs. |
Limitations
An undeployment plan can only include objects from a single workspace.
A plan can contain a maximum of 200 objects.
You can roll back task operations before the Undeploy Task step and table operations before the Undeploy Completed step.
Permissions
To create an undeployment plan, you must have one of the following permissions:
Alibaba Cloud account
A RAM user with the
AliyunDataWorksFullAccesspermissionworkspace administrator
tenant-level data governance administrator
workspace-level data governance administrator
O&M engineer
Only an Alibaba Cloud account, a RAM user with the
AliyunDataWorksFullAccesspermission, or a workspace administrator can perform all operations on an undeployment plan.O&M engineers can manage only the undeployment plans they create. They have read-only permissions for other plans.
Tenant-level and workspace-level data governance administrators can only create, configure, and delete plans. They cannot start or execute plans that contain tasks for which they do not have permissions.
Usage notes
An object already in an undeployment plan cannot be added to another.
Certain types of objects cannot be undeployed, such as the workspace root node or virtual nodes created by an undeployment plan.
If owners do not remove scheduling dependencies from an object within the specified time, the system creates a virtual node and attaches the dependent tasks to it.
Supported steps may vary by node type. Refer to the UI for details.
You can only roll back operations performed before the object is undeployed.
Procedure
Step 1: Create an undeployment plan
Go to the Data Asset Governance page.
Log on to the DataWorks console. In the top navigation bar, select the desired region. In the left-side navigation pane, choose . On the page that appears, click Go to Data Asset Governance.
In the left-side navigation pane, choose to go to the Graceful Shutdown page.
From the Workspace drop-down list at the top of the page, select the workspace that contains the objects that you want to undeploy. The list shows only workspaces that you have joined.
If you need to join a workspace, contact the workspace administrator to add you. For more information, see Add workspace members and manage their roles and permissions.
On the Graceful Shutdown page, click Create Governance Plan.
Alternatively, on the Process Governance Items page, find the task or table that you want to undeploy and click Create Plan in the Actions column.
Step 2: Select objects to undeploy
1. Configure basic plan information
Specify the Workspace where the objects are located and enter a custom Plan Name.
2. Add objects to the plan
In the To-Be-Undeployed Object section, select the objects to undeploy. You can add published Tasks or Table.
Undeploy tasks: You can add tasks by using Add to-be-undeployed Object or Import Based on Lineage.
Add to-be-undeployed Object: Add task objects by node ID.
Import Based on Lineage: Select a task and then select related tasks based on their data lineage hierarchy.
NoteYou can find the node ID for a task on its configuration page under Scheduling > Base Attributes.
You can also find the node ID on the page in Operation Center.
Undeploy tables: Use Add to-be-undeployed Object to add tables.
An object that is already part of an undeployment plan cannot be added to another one.
Certain types of objects cannot be undeployed, such as the workspace root node or virtual nodes created by an undeployment plan.
Parameter | Description |
Undeployment Impact Assessment |
|
Impact Item Description | Provides an explanation of the undeployment impact assessment. |
The output table is consumed. | The consumption status of the task object's output table. |
Task That Generates Table | The task that generates the table object. Click to view task information. |
Number of Affected Baselines/Highest Priority |
Note For more information about task baseline monitoring, see Baseline management. |
Step 3: Configure the undeployment process
After adding the objects, configure the multi-stage undeployment process. The process can include up to eight stages: Start Undeployment, Delayed Scheduling, Pause scheduling, Undeploy Task, Rename Output Table, Reset Table Lifecycle, Delete Output Table, and Undeploy Completed.
Stage | Description | Recommended use case |
Start Undeployment | Starts the graceful undeployment process. This stage is mandatory. | - |
Delayed Scheduling | Delays the object's scheduled run time, which in turn delays data generation for it and its downstream dependencies. | Use delayed scheduling to ensure that tasks run normally when data from upstream platforms is not yet generated. |
Pause scheduling | Sets the status of the object to frozen. This means the task's status appears as frozen on the page. The object and its downstream dependencies will no longer generate table data. | Use suspend scheduling when a business service is paused but the workflow needs to be retained. |
Undeploy Task | Undeploys the selected object. No new instances are generated starting the next day. Existing instances are set to zero load and will no longer run. This stage is selected by default but can be deselected. | Use this when a node or workflow is obsolete. |
Rename Output Table | Renames the output table of the task or table object. | Use this to mark the output table as belonging to an undeployed object. |
Reset Table Lifecycle | Resets the lifecycle of the output table for the task or table object. | Use this to retain the output table for a specific period after undeployment. Note Only MaxCompute tables are supported. |
Delete Output Table | Deletes the output table of the task or table object. | Use this after an object is undeployed and its output table is no longer needed. |
Undeploy Completed | Ends the graceful undeployment process. This stage is mandatory. | - |
Stages are executed sequentially. Optional stages can be skipped.
Graceful undeployment plans support sending notifications to specified personnel through methods such as System Prompt, Mail, and DingTalk Chatbot. For information about how to configure a DingTalk chatbot, see Scenario practice: Send alert notifications to a DingTalk group. To view system prompts, go to the Governance Reminder page by choosing .
Step 4: Define the execution mode
Set how the plan progresses through each stage. You can choose Manually Trigger Execution or Automatic Execution.
Manually Trigger Execution
Description: The plan creator must manually confirm each stage before the process can continue.
Recommended use case: Use this mode when the task or data table is critical to your business workflow.
Automatic Execution
Description: Each step proceeds automatically when the configured quiet period ends.
Recommended use case: Use this mode to undeploy numerous low-priority workflows in bulk.
Step 5: View and manage the plan
On the Graceful Shutdown page, you can view the list of undeployment plans and perform actions on them, such as Execute, Cancel, or Delete.
Canceling a plan only stops in-progress stages; completed stages are unaffected.
Deleting a plan first cancels any in-progress stages, then deletes the plan itself. Completed stages are unaffected. This action is irreversible.
Click Execution details in the Actions column to view a plan's configuration and manage its individual tasks:
Function | Description |
Details | View the undeployment details for a single task, including progress, status, and execution details for each stage. You can also cancel or roll back the undeployment process for the current task.
|
Go to Operation Center | Opens the task's details in Operation Center. |
Edit | Opens the task for editing in DataStudio. |