如何基于Service Bus Trigger的deliveryCount设置Polly重试延迟

方案一：函数内部动态生成重试策略 #

直接在Azure函数方法内，利用拿到的deliveryCount参数生成对应的Polly重试策略，无需在DI阶段提前配置固定策略。

首先将函数改为异步（适配Polly异步策略与HttpClient异步调用），注入IHttpClientFactory，代码示例如下：

[FunctionName("ServiceBusQueueTriggerCSharp")]                    
public static async Task Run(
    [ServiceBusTrigger("myqueue", Connection = "ServiceBusConnection")] 
    string myQueueItem,
    Int32 deliveryCount,
    IHttpClientFactory httpClientFactory,
    ILogger log)
{
    var httpClient = httpClientFactory.CreateClient(ClientNames.HttpClient);
    
    // 根据当前消息的deliveryCount生成重试延迟规则
    var retryPolicy = HttpPolicyExtensions
        .HandleTransientHttpError()
        .OrResult(msg => msg.StatusCode == System.Net.HttpStatusCode.NotFound)
        .WaitAndRetryAsync(1, _ => TimeSpan.FromSeconds(5 * deliveryCount)); // 用deliveryCount替代retryAttempt计算延迟

    try
    {
        await retryPolicy.ExecuteAsync(async () => 
        {
            var response = await httpClient.GetAsync("目标API地址");
            response.EnsureSuccessStatusCode();
            return response;
        });
    }
    catch (Exception ex)
    {
        log.LogError(ex, "HTTP调用经重试后仍失败");
        throw; // 抛出异常让ServiceBus按配置处理消息重试
    }
}

方案二：通过Polly Context传递deliveryCount #

保留DI中的策略框架，通过Polly的Context对象将deliveryCount动态传递给全局策略，在策略中读取该值计算延迟。

第一步：修改DI中的策略配置

services.AddHttpClient(ClientNames.HttpClient)
    .SetHandlerLifetime(TimeSpan.FromMinutes(5))
    .AddPolicyHandler(GetRetryPolicyWithContext());

private static IAsyncPolicy<HttpResponseMessage> GetRetryPolicyWithContext()
{
    return HttpPolicyExtensions
        .HandleTransientHttpError()
        .OrResult(msg => msg.StatusCode == System.Net.HttpStatusCode.NotFound)
        .WaitAndRetryAsync(1, (_, context) => 
        {
            // 从Context中读取deliveryCount，读取失败则用默认值兜底
            if (context.TryGetValue("DeliveryCount", out var deliveryCountObj) && deliveryCountObj is int deliveryCount)
            {
                return TimeSpan.FromSeconds(5 * deliveryCount);
            }
            return TimeSpan.FromSeconds(5);
        });
}

第二步：函数中传递Context调用HttpClient

[FunctionName("ServiceBusQueueTriggerCSharp")]                    
public static async Task Run(
    [ServiceBusTrigger("myqueue", Connection = "ServiceBusConnection")] 
    string myQueueItem,
    Int32 deliveryCount,
    IHttpClientFactory httpClientFactory,
    ILogger log)
{
    var httpClient = httpClientFactory.CreateClient(ClientNames.HttpClient);
    // 创建Polly Context并传入deliveryCount
    var pollyContext = new Context { { "DeliveryCount", deliveryCount } };

    try
    {
        // 必须使用SendAsync并传入Context，确保策略能读取到传递的参数
        var response = await httpClient.SendAsync(new HttpRequestMessage(HttpMethod.Get, "目标API地址"), pollyContext);
        response.EnsureSuccessStatusCode();
    }
    catch (Exception ex)
    {
        log.LogError(ex, "HTTP调用经重试后仍失败");
        throw;
    }
}

注意事项 #

• 两种方案都建议将函数改为异步，避免阻塞Azure函数宿主，同时适配Polly与HttpClient的异步API。
• 需注意Polly的重试逻辑与ServiceBus自身的消息重试（如MaxDeliveryCount配置）的层级关系，避免重复重试导致不必要的延迟。
• 方案一更适合无需复用策略结构的场景，实现更直接；方案二更适合需要复用策略逻辑，仅动态调整延迟参数的场景。

内容的提问来源于stack exchange，提问作者A B