Function Compute:Asynchronous tasks

Limitations

• Use case limitations

  Asynchronous tasks offer more features but also introduce system overhead. We recommend disabling task mode in the following scenarios:

  • Your application is highly sensitive to end-to-end request latency and requires an average latency of less than 100 ms.

  • You need to make asynchronous invocations at thousands of requests per second (RPS) or higher.

• Region limitations

  Asynchronous tasks are supported in the following regions: China (Hangzhou), China (Shanghai), China (Qingdao), China (Beijing), China (Zhangjiakou), China (Ulanqab), China (Shenzhen), China (Chengdu), China (Hong Kong), Singapore, UK (London), US (Silicon Valley), US (Virginia), Germany (Frankfurt), Malaysia (Kuala Lumpur), Indonesia (Jakarta), Thailand (Bangkok), Japan (Tokyo), and South Korea (Seoul).

• Retention period

  You can query the status of tasks created within the last 7 days.

Comparison with Kubernetes jobs

To run asynchronous tasks, you can use the asynchronous task feature of Function Compute or Kubernetes jobs. Function Compute provides a ready-to-use, serverless platform for running asynchronous tasks that is more efficient, controllable, and observable. The following table compares the two methods.

Task management

Asynchronous task statuses

For each task invocation, Function Compute records state transitions and enables real-time status queries. You can view the specific status of a task by using an SDK or the Function Compute console. A task can be in one of the following statuses:

Call the GetAsyncTask API operation to get details about a task execution, or call the ListAsyncTasks API operation to filter tasks by status.

Manage tasks

{
  "asyncTask": true
}

Event triggers

You can trigger asynchronous tasks by creating a trigger that supports asynchronous invocation. The following triggers support this feature:

• HTTP trigger

• Time trigger

• OSS trigger

• MNS topic trigger

• Triggers integrated with EventBridge, including:

  • MNS queue trigger

  • RocketMQ trigger

  • RabbitMQ trigger

  • Kafka trigger

  • MQTT trigger

  • DTS trigger

  Important

  Make sure that the Invocation Method for HTTP triggers and triggers integrated with EventBridge is set to Async Invocation.

Task deduplication

To prevent duplicate executions, Function Compute lets you assign a globally unique ID to each task. If a task submission fails, for example due to a timeout, you can safely retry the request by using the same task ID. For more information about retry behavior, see retry policy.

Function Compute uses a globally unique TaskID for deduplication. We recommend specifying a TaskID for every task submission and use it for retries in case of timeouts or other failures. Function Compute rejects subsequent submissions with a duplicate TaskID with a 409 error.

Set a TaskID

You can submit a task from the Function Compute console, by using Serverless Devs, or by calling an API operation. To set a TaskID, specify the x-fc-async-task-id header in the invocation request.

For sample code on how to set a TaskID when triggering a task, see Sample Code (International site).

Task monitoring

You can use the monitoring metrics provided by Function Compute to get real-time status updates for asynchronous tasks and promptly detect errors, timeouts, and other exceptions.

Prerequisites

A function is created. For more information, see Create a function.

Procedure

1. Log on to the Function Compute console. In the left-side navigation pane, click Functions.

2. In the top navigation bar, select a region. On the Function page, click the target function.

3. On the function details page, click the Monitoring tab to view metrics related to asynchronous invocations and resource usage for asynchronous task instances. For more information about monitoring metrics, see Monitoring metrics.

   • Click the Function Metrics tab to view metrics related to asynchronous invocations, such as Asynchronous Invocation Processing (count), Asynchronous Invocation Processing Latency (ms), Asynchronous Invocation Trigger Events (count), and Asynchronous Requests Backlogs (count).

   • Click the Instance Metrics tab to view instance-level resource usage for asynchronous tasks.

     Note

     Before you can view instance-level metrics, you must enable the logging feature. For more information, see Configure logging.

4. Set an alert rule for a metric.

   To monitor a metric such as memory usage or vCPU usage, click the bell icon in the upper-right corner of the metric chart. You will be redirected to the Create Metric-based Alert Rule panel. Set the alert conditions and threshold as prompted, and then click OK.

   The following figure shows an example of an alert rule configuration. In this example, if the number of backlogged asynchronous invocations for a function is greater than or equal to 5, a medium-severity alert is sent 24/7 to the specified contacts via DingTalk, email, phone, and webhook.

   

Task orchestration

Integrate with CloudFlow to use asynchronous function invocations in large-scale, complex task workflows. You can use states like Pass, Map, and Parallel to orchestrate Function Compute asynchronous tasks. CloudFlow reliably coordinates executions according to predefined steps, tracks each task's state transitions, and applies your retry logic as needed to ensure the workflow completes successfully. For more information, see Integrate with Function Compute.

Prerequisites

• Create an asynchronous task. For more information, see Task management.

• Grant CloudFlow access to Function Compute. For more information, see Use a function role to grant Function Compute permissions to access other cloud services.

Procedure

1. Log on to the CloudFlow console. In the top navigation bar, select a region.

2. In the left-side navigation pane, click Workflows. On the Workflows page, click Create Workflow.

3. In the Create Workflow dialog box, select a method and a mode for creating the workflow.

   CloudFlow supports two workflow modes: Standard and Express. For more information about the two modes, see Standard workflows and Express workflows. For more information about creating a workflow, see Create a workflow from a blank canvas.

4. Edit the workflow you just created.

   1. Define the flow.

      You can use the CloudFlow Studio editor or the YAML editor. This example uses the YAML editor.

      In the upper-right corner of the flow details page, click YAML. Enter the flow definition language in the code editor on the left, and then click Save.

      The following is an example:

      Type: StateMachine
      Name: MyFlow
      SpecVersion: v1
      StartAt: InvokeFunction
      States:
        - Type: Task
          Name: InvokeFunction
          Action: FC:InvokeFunction
          TaskMode: RequestComplete
          Parameters:
            invocationType: Async
            resourceArn: acs:fc:us-west-1:103435468****:functions/func-i1****/LATEST
          End: true

   2. Configure the execution role.

      In the upper-right corner of the flow details page, click Configure Workflow. Select an Execution Role, and then click Save.

      Note

      Ensure the role you configure for the flow is attached to the AliyunFCInvocationAccess policy.

5. On the flow details page, you can click Execute in the upper-right corner to test the workflow.