你当前正在访问 Microsoft Azure Global Edition 技术文档网站。 如果需要访问由世纪互联运营的 Microsoft Azure 中国技术文档网站，请访问 https://docs.azure.cn。

Durable Functions 中的子业务流程 (Azure Functions)

除了调用活动函数之外，业务流程协调程序函数还可以调用其他业务流程协调程序函数。 例如，可以从较小的业务流程协调程序函数库构建较大的业务流程。 或者，你可以并行运行某个业务流程协调程序函数的多个实例。

业务流程协调程序函数可以使用 call-sub-orchestrator API 调用另一个业务流程协调程序函数。 错误处理和修正一文提供了有关自动重试的更多信息。

从调用方的角度来看，子业务流程协调程序函数的行为与活动函数相同。 它们可以返回值，引发异常，并且父业务流程协调程序函数可以等待它们。

注意

PowerShell 尚不支持子业务流程。

注意

适用于 Azure Functions 的 Node.js 编程模型版本 4 现已正式发布。 新版 v4 模型旨在为 JavaScript 和 TypeScript 开发人员提供更为灵活和直观的体验。 在迁移指南中详细了解 v3 和 v4 之间的差异。

在以下代码片段中，JavaScript (PM4) 表示编程模型 V4，即新体验。

示例

以下示例说明了一个需要预配多台设备的 IoT（物联网）方案。 以下函数表示需要为每个设备执行的预配工作流：

C# (InProc)

public static async Task DeviceProvisioningOrchestration(
    [OrchestrationTrigger] IDurableOrchestrationContext context)
{
    string deviceId = context.GetInput<string>();

    // Step 1: Create an installation package in blob storage and return a SAS URL.
    Uri sasUrl = await context.CallActivityAsync<Uri>("CreateInstallationPackage", deviceId);

    // Step 2: Notify the device that the installation package is ready.
    await context.CallActivityAsync("SendPackageUrlToDevice", Tuple.Create(deviceId, sasUrl));

    // Step 3: Wait for the device to acknowledge that it has downloaded the new package.
    await context.WaitForExternalEvent<bool>("DownloadCompletedAck");

    // Step 4: ...
}

C#（独立）

public static async Task DeviceProvisioningOrchestration(
    [OrchestrationTrigger] TaskOrchestrationContext context, string deviceId)
{
    // Step 1: Create an installation package in blob storage and return a SAS URL.
    Uri sasUrl = await context.CallActivityAsync<Uri>("CreateInstallationPackage", deviceId);

    // Step 2: Notify the device that the installation package is ready.
    await context.CallActivityAsync("SendPackageUrlToDevice", (deviceId, sasUrl));

    // Step 3: Wait for the device to acknowledge that it has downloaded the new package.
    await context.WaitForExternalEvent<bool>("DownloadCompletedAck");

    // Step 4: ...
}

JavaScript (PM3)

const df = require("durable-functions");

module.exports = df.orchestrator(function*(context) {
    const deviceId = context.df.getInput();

    // Step 1: Create an installation package in blob storage and return a SAS URL.
    const sasUrl = yield context.df.callActivity("CreateInstallationPackage", deviceId);

    // Step 2: Notify the device that the installation package is ready.
    yield context.df.callActivity("SendPackageUrlToDevice", { id: deviceId, url: sasUrl });

    // Step 3: Wait for the device to acknowledge that it has downloaded the new package.
    yield context.df.waitForExternalEvent("DownloadCompletedAck");

    // Step 4: ...
});

JavaScript (PM4)

const df = require("durable-functions");

df.app.orchestration("deviceProvisioningOrchestration", function* (context) {
    const deviceId = context.df.getInput();

    // Step 1: Create an installation package in blob storage and return a SAS URL.
    const sasUrl = yield context.df.callActivity("createInstallationPackage", deviceId);

    // Step 2: Notify the device that the installation package is ready.
    yield context.df.callActivity("sendPackageUrlToDevice", { id: deviceId, url: sasUrl });

    // Step 3: Wait for the device to acknowledge that it has downloaded the new package.
    yield context.df.waitForExternalEvent("downloadCompletedAck");

    // Step 4: ...
});

Python

import azure.functions as func
import azure.durable_functions as df

def orchestrator_function(context: df.DurableOrchestrationContext):
    device_id = context.get_input()

    # Step 1: Create an installation package in blob storage and return a SAS URL.
    sas_url = yield context.call_activity("CreateInstallationPackage", device_id)

    # Step 2: Notify the device that the installation package is ready.
    yield context.call_activity("SendPackageUrlToDevice", { "id": device_id, "url": sas_url })

    # Step 3: Wait for the device to acknowledge that it has downloaded the new package.
    yield context.call_activity("DownloadCompletedAck")

    # Step 4: ...

Java

@FunctionName("DeviceProvisioningOrchestration")
public void deviceProvisioningOrchestration(
        @DurableOrchestrationTrigger(name = "ctx") TaskOrchestrationContext ctx) {
    // Step 1: Create an installation package in blob storage and return a SAS URL.
    String deviceId = ctx.getInput(String.class);
    String blobUri = ctx.callActivity("CreateInstallPackage", deviceId, String.class).await();

    // Step 2: Notify the device that the installation package is ready.
    String[] args = { deviceId, blobUri };
    ctx.callActivity("SendPackageUrlToDevice", args).await();

    // Step 3: Wait for the device to acknowledge that it has downloaded the new package.
    ctx.waitForExternalEvent("DownloadCompletedAck").await();

    // Step 4: ...
}

此业务流程协调程序函数可以按现样用于一次性设备预配，也可以用作大型业务流程的一部分。 在后一种情况下，父业务流程协调程序函数可以使用 call-sub-orchestrator API 调度 DeviceProvisioningOrchestration 的实例。

下面是一个示例，它展示了如何并行运行多个业务流程协调程序函数。

C# (InProc)

[FunctionName("ProvisionNewDevices")]
public static async Task ProvisionNewDevices(
    [OrchestrationTrigger] IDurableOrchestrationContext context)
{
    string[] deviceIds = await context.CallActivityAsync<string[]>("GetNewDeviceIds");

    // Run multiple device provisioning flows in parallel
    var provisioningTasks = new List<Task>();
    foreach (string deviceId in deviceIds)
    {
        Task provisionTask = context.CallSubOrchestratorAsync("DeviceProvisioningOrchestration", deviceId);
        provisioningTasks.Add(provisionTask);
    }

    await Task.WhenAll(provisioningTasks);

    // ...
}

注意

前面的 C# 示例适用于 Durable Functions 2.x。 对于 Durable Functions 1.x，必须使用 DurableOrchestrationContext 而不是 IDurableOrchestrationContext。 有关版本之间差异的详细信息，请参阅 Durable Functions 版本一文。

C#（独立）

[FunctionName("ProvisionNewDevices")]
public static async Task ProvisionNewDevices(
    [OrchestrationTrigger] TaskOrchestrationContext context)
{
    string[] deviceIds = await context.CallActivityAsync<string[]>("GetNewDeviceIds");

    // Run multiple device provisioning flows in parallel
    var provisioningTasks = new List<Task>();
    foreach (string deviceId in deviceIds)
    {
        Task provisionTask = context.CallSubOrchestratorAsync("DeviceProvisioningOrchestration", deviceId);
        provisioningTasks.Add(provisionTask);
    }

    await Task.WhenAll(provisioningTasks);

    // ...
}

JavaScript (PM3)

const df = require("durable-functions");

module.exports = df.orchestrator(function*(context) {
    const deviceIds = yield context.df.callActivity("GetNewDeviceIds");

    // Run multiple device provisioning flows in parallel
    const provisioningTasks = [];
    var id = 0;
    for (const deviceId of deviceIds) {
        const child_id = context.df.instanceId+`:${id}`;
        const provisionTask = context.df.callSubOrchestrator("DeviceProvisioningOrchestration", deviceId, child_id);
        provisioningTasks.push(provisionTask);
        id++;
    }

    yield context.df.Task.all(provisioningTasks);

    // ...
});

JavaScript (PM4)

const df = require("durable-functions");

df.app.orchestration("provisionNewDevices", function* (context) {
    const deviceIds = yield context.df.callActivity("getNewDeviceIds");

    // Run multiple device provisioning flows in parallel
    const provisioningTasks = [];
    var id = 0;
    for (const deviceId of deviceIds) {
        const child_id = context.df.instanceId + `:${id}`;
        const provisionTask = context.df.callSubOrchestrator(
            "deviceProvisioningOrchestration",
            deviceId,
            child_id
        );
        provisioningTasks.push(provisionTask);
        id++;
    }

    yield context.df.Task.all(provisioningTasks);

    // ...
});

Python

import azure.functions as func
import azure.durable_functions as df

def orchestrator_function(context: df.DurableOrchestrationContext):

    device_IDs = yield context.call_activity("GetNewDeviceIds")

    # Run multiple device provisioning flows in parallel
    provisioning_tasks = []
    id_ = 0
    for device_id in device_IDs:
        child_id = f"{context.instance_id}:{id_}"
        provision_task = context.call_sub_orchestrator("DeviceProvisioningOrchestration", device_id, child_id)
        provisioning_tasks.append(provision_task)
        id_ += 1

    yield context.task_all(provisioning_tasks)

    # ...

Java

@FunctionName("ProvisionNewDevices")
public void provisionNewDevices(
        @DurableOrchestrationTrigger(name = "ctx") TaskOrchestrationContext ctx) {
    List<?> deviceIDs = ctx.getInput(List.class);

    // Schedule each device provisioning sub-orchestration to run in parallel
    List<Task<Void>> parallelTasks = deviceIDs.stream()
        .map(device -> ctx.callSubOrchestrator("DeviceProvisioningOrchestration", device))
        .collect(Collectors.toList());

    // ...
}

注意

子业务流程必须在与父业务流程相同的函数应用中定义。 如果需要在另一个函数应用中调用并等待业务流程，请考虑使用对 HTTP API 和 HTTP 202 轮询使用者模式的内置支持。 有关详细信息，请参阅 HTTP 功能主题。

后续步骤

了解如何设置自定义业务流程状态