Platform For AI:PAI workspace event notification configuration

1.1 Activate EventBridge

On the Event Notification Configuration tab, click Activate for Free to go to the EventBridge activation page. For more information, see Activate EventBridge.

After activation, PAI automatically creates a custom event bus named pai-system-${workspace_name} for each PAI workspace. You can go to the EventBridge console, switch to the region where your workspace resides, and view and manage your custom event buses.

1.2 Grant PAI access to cloud resources

Click Authorize Now. The system automatically creates the AliyunServiceRoleForPAIWorkspace service-linked role, which authorizes PAI to access your cloud resources. For more information about this role, see Appendix: PAI workspace service-linked role.

In the one-click authorization dialog box that appears, click Authorize.

1.3 (Optional) Grant event notification permissions to a RAM user

If you use a Resource Access Management (RAM) user (sub-account) for configuration, grant that user permissions to manage the event bus. Otherwise, the RAM user cannot create or manage event rules.

Create a custom policy and attach it to the RAM user. The policy content is as follows:

{
  "Statement": [{
    "Effect": "Allow",
    "Action": [
      "eventbridge:CreateEventBus",
      "eventbridge:GetEventBus",
      "eventbridge:DeleteEventBus",
      "eventbridge:ListEventBuses",
      "eventbridge:CreateRule",
      "eventbridge:GetRule",
      "eventbridge:UpdateRule",
      "eventbridge:EnableRule",
      "eventbridge:DisableRule",
      "eventbridge:DeleteRule",
      "eventbridge:ListRules",
      "eventbridge:PutEvents",
      "eventbridge:UpdateTargets",
      "eventbridge:DeleteTargets",
      "eventbridge:ListTargets"
    ],
    "Resource": "acs:eventbridge:*:*:eventbus/*"
  }],
  "Version": "1"
}

2.1 Configure event type

Select the Event Source and the specific Event Type that you want to monitor. The following event sources are supported:

• Pipeline Jobs: Monitor the status of Designer pipeline tasks. Event types include:

  • Job Failure

  • Successful

• DLC Jobs: Monitor the lifecycle events of a DLC job. You can select multiple event types. Event types include:

  • Job Progress: Enter Queue, Start Bidding, Start Environment Preparation, Start Run, Retained upon Success, Retained upon Failure, Job Failure, and Successful.

  • Automatic Fault Tolerance

  • Job Timeout: Environment Preparation Timeout, Wait Timeout, Queue Timeout, and Run Timeout. If you select a timeout event type, configure the timeout rule in scheduling configuration first.

  • Other Events: Job Preempted (only idle jobs and spot jobs can be preempted), Job Manually Stopped, and Job Priority Modified.

• Models: Monitor the status changes of a model version in AI Asset Management. Event types include:

  • Version Approved: The model version is approved for deployment.

  • Model version status change (approved or rejected for deployment).

• DSW instance: Monitor the status changes of a DSW instance and its saved runtime images. You can select multiple event types. Event types include:

  • Status history: Queuing, Running, Instance deleted, Updating, Failed, Allocating, Recovering, Starting, and Preparing environment.

  • Save image: Saving image, Image saved, Failed to save image, and Image saving timeout.

  • Stop instance: Stop instance and Stopping.

2.2 Configure event targets

Configure how and where to deliver notifications when an event occurs. Note that the HTTP/HTTPS target applies only to Model event types.

• DingTalk Notification: Configure the webhook and signing key parameters. For more information, see Appendix: Webhook and signing key. After you configure the parameters, click Test Connectivity to run a connectivity validation.

• WeCom Notification: Configure the webhook URL for WeCom messages.

• Lark Notification: Configure the webhook URL for a custom Lark bot.

• HTTP/HTTPS: Automatically calls a specified URL when the status of a model version changes. This target applies only to Model event types, and your receiving API must parse the request based on the required template.

By default, each event rule supports up to 5 event targets. If this quota does not meet your requirements, request a quota increase. A maximum of 100 is recommended.