Azure Durable Functions永恒编排异常问题排查与解决

问题场景 #

尝试基于Azure Durable Functions实现永恒编排（Eternal Orchestration），期望编排能循环执行，但运行时抛出IllegalStateException: The orchestrator has already completed异常，无法正常循环。

相关代码 #

HTTP Trigger 代码 #

@FunctionName("MQMessageOrchestrate")
public HttpResponseMessage mqMessageOrchestrate(
        @HttpTrigger(name = "req", methods = { HttpMethod.GET,
                HttpMethod.POST }, authLevel = AuthorizationLevel.ANONYMOUS) HttpRequestMessage<Optional<String>> request,
        @DurableClientInput(name = "durableContext") DurableClientContext durableContext,
        final ExecutionContext context) {

    context.getLogger().info("Java HTTP trigger processed a request.");

    // Initialize durable task client
    DurableTaskClient client = durableContext.getClient();

    // Start orchestration
    String instanceId = client.scheduleNewOrchestrationInstance("TestOrchestrator");

    // Output instance id
    context.getLogger().info("Created new Java orchestration with instance ID = " + instanceId);

    // Return status of orchestration in response
    return durableContext.createCheckStatusResponse(request, instanceId);
}

初始Orchestrator代码（异常版本） #

@FunctionName("TestOrchestrator")
public void testOrchestrator(
        @DurableOrchestrationTrigger(name = "ctx") TaskOrchestrationContext ctx, ExecutionContext context) {

    System.out.println("Processing stuff...");

    ctx.createTimer(Duration.ofSeconds(5)).await();

    ctx.continueAsNew(null);

}

异常信息 #

• 预期行为：编排循环执行
• 实际异常：IllegalStateException: The orchestrator has already completed
• 错误日志：

[2024-12-05T01:54:06.693Z] Processing stuff...
[2024-12-05T01:54:06.714Z] Executed 'Functions.TestOrchestrator' (Failed, Id=b8aff574-c31f-4e7a-8269-571d303ec402, Duration=27ms)
[2024-12-05T01:54:06.715Z] Exception: IllegalStateException: The orchestrator has already completed

解决方案 #

将timer操作移至continueAsNew调用之后，修改后的Orchestrator代码可正常实现永恒编排：

@FunctionName("TestOrchestrator")
public void testOrchestrator(
        @DurableOrchestrationTrigger(name = "ctx") TaskOrchestrationContext ctx, ExecutionContext context) {

    ctx.callActivity("DoCleanup", "test").await();

    context.getLogger().info("Orchestration restarted as new ...");

    ctx.continueAsNew(null);

    ctx.createTimer(Duration.ofSeconds(0)).await();
}

原因说明 #

continueAsNew调用后，当前编排实例会被标记为已完成状态。如果在continueAsNew之前执行await操作，Durable Functions运行时会认为编排已经结束，后续的状态变更会触发异常。将立即执行的timer放在continueAsNew之后并await，可以让运行时正确处理实例过渡，避免触发状态检查异常，从而实现永恒编排的循环执行。

内容的提问来源于stack exchange，提问作者Carlos Iñigo