Function Compute:Configure instance lifecycles

Function instance lifecycle

Function instances are dynamically created and destroyed based on the current request volume. The lifecycle of each function instance includes three phases: Creating, Invoke, and Destroy, as shown in the following figure.

Creating

Instance creation refers to Function Compute creating function instances based on your function configuration. During the instance creation phase, Function Compute performs the following three tasks in sequence:

1. Instance Create: Function Compute loads code and layers, or pulls images, and then initializes an instance.

2. Runtime Init: Function Compute initializes the runtime.

3. Init Hook: Function Compute executes the configured Initializer hook. For more information, see Initializer hooks.

The following figure shows the Creating phase.

Invoke

During instance runtime, your function handler is called to process function invocation requests from internal and external sources. In the Invoke phase, for built-in runtimes supported by Function Compute, an instance processes only one request at a time. For custom runtimes or custom container runtimes, an instance can process multiple requests simultaneously. You can enable multiple concurrent requests per instance by configuring instance concurrency. For more information, see Configure instance concurrency.

Function Compute charges only for actual requests and hook executions. Instances are frozen during periods without requests and therefore not billed. For more information, see Billing.

Destroy

This phase is triggered if a function instance does not receive any invocations for a period of time. During the Destroy phase, Function Compute first executes the PreStop hook. You can perform cleanup tasks in PreStop hooks.

Instances are destroyed in the following scenarios:

• Idle instances: If an instance does not receive any invocation requests for a period of time, Function Compute automatically reclaims the instance.

• Minimum instance adjustment: When you reduce the number of minimum instances, Function Compute immediately destroys the excess instances.

• Instance exceptions: If an instance encounters an exception during the Creating or Running phase, Function Compute destroys the instance.

Instance freezing mechanism

When there are no invocation requests, Function Compute freezes (Freeze) the instance, and when new requests arrive, Function Compute thaws (Thaw) the instance. The following figure shows this process.

An instance is frozen in the following scenarios:

• The instance is initialized but not yet invoked by any requests.

• The instance is invoked to process requests and subsequent requests are not immediately submitted.

After an invocation phase is completed, Function Compute freezes the function instance. Background processes, threads, or coroutines in the program cannot continue running, and asynchronous logs may not be written successfully.

Limits

• GPU functions support two types of Initializer hooks: Code Invocation and Command Execution. These two types cannot be configured simultaneously. Only one can be active.

• For event, Web, and task functions, only the Code Invocation type is supported. This type is enabled by default, and you do not need to select it manually.

• PreStop hooks only support the Code Invocation type, and all runtimes support PreStop hooks.

• The input parameters of a PreStop hook do not contain an event parameter.

• The PreStop hook does not support returning values, and any logic added to the hook to return values is ineffective.

• If you use a Java Runtime, you must update fc-java-core to 1.4.0 or later to utilize the PreStop hook functionality.

• When a function returns responses, Function Compute freezes the function instance. You cannot assume that all asynchronous processes, threads, and coroutines are successfully executed when responses are returned. You cannot assume that asynchronous logs are refreshed either.

• For Code Invocation type hooks, the configuration methods differ across runtimes. For built-in runtimes, you need to customize the hook entry point when configuring lifecycle hooks. For example, if you configure the Initializer hook as index.initialize, you need to add an initialize callback function in your code. For custom runtimes and custom container runtimes, after configuring Initializer and PreStop hooks, the system sends HTTP requests (POST /initialize or GET/pre-stop) to your function when the instance starts or stops. You need to respond to these requests in your business code.

Billing

You are not charged for the number of invocations for instance lifecycle hooks. Other billable items are the same as regular invocations. The following figure shows the billing durations. For more information about billing methods, see Billing overview.

Requirement

Function creation is completed. For more information, see Create a function.

Configure instance lifecycle hooks

  codeUri: './code.zip'
  ......
  instanceLifecycleConfig:
    initializer:
      timeout: 60
      command:
        - /bin/sh
        - -c
        - echo "hello"

  codeUri: './code.zip'
  ......
  instanceLifecycleConfig:
    initializer:
      handler: ""
      timeout: 60
      command: ""

Query hook-related logs

After you configure an instance lifecycle hook and run its code, you can query the logs related to the hook.

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 Functions page, click the function that you want to manage.

3. On the Function Details page, click the Logs tab. On the Invocation Requests tab, find the request whose logs you want to view and click Actions column's Advanced Logs.

   You can use a copied instance ID to query the start and end logs of all lifecycle hooks. You can also use instance ID + lifecycle hook keyword to query the start and end logs of a specific hook, for example, c-62833f38-20f1629801fa4bd***** and PreStop.

   You can also query the request logs based on the request ID in the start and end logs. If the user log does not contain a request ID, you can click the icon to obtain the context log.