网站导航
> 技术文档 > 【进阶】C#.NET 处理瞬态故障 polly库 详解_c# polly

【进阶】C#.NET 处理瞬态故障 polly库 详解_c# polly

网友: 技术文档 2025-07-30 10:30:51

简介

Polly 是处理 .NET 应用中瞬态故障的强大工具，通过提供丰富的弹性策略，帮助开发者构建更健壮、更具弹性的应用程序。合理使用 Polly 可以显著提高应用在面对网络波动、服务临时不可用等问题时的稳定性。

Polly 核心策略详解

Polly 提供七种策略应对不同故障场景，均支持流畅 API 链式调用

  • 重试 (Retry)：短暂故障（如网络抖动）

  • 熔断 (Circuit Breaker)：持续故障避免系统过载

  • 超时 (Timeout)：防止长时间阻塞

  • 隔板隔离 (Bulkhead)：限制并发资源消耗

  • 回退 (Fallback)：失败时提供降级响应

  • 缓存 (Cache)：高频数据加速访问

  • 策略包装 (PolicyWrap)：组合多策略应对复杂故障

应用场景：

  • 调用远程服务（REST API、微服务）

  • 数据库操作

  • 文件系统访问

  • 任何可能出现暂时性故障的操作

示例 - 熔断+重试组合策略

var policy = Policy.Wrap( Policy.Handle<HttpRequestException>() .CircuitBreaker(5, TimeSpan.FromSeconds(30)), Policy.Handle<TimeoutException>() .WaitAndRetry(3, attempt => TimeSpan.FromSeconds(attempt)));policy.Execute(() => CallExternalService());

Polly 基本使用

安装 NuGet 包
dotnet add package Polly
基本策略使用模式
// 创建策略var policy = Policy .Handle<HttpRequestException>() // 处理特定异常 .Retry(3); // 重试3次// 执行受策略保护的操作await policy.ExecuteAsync(async () =>{ // 可能会失败的操作（如HTTP请求） var response = await httpClient.GetAsync(\"https://api.example.com/data\"); response.EnsureSuccessStatusCode(); return await response.Content.ReadAsStringAsync();});

核心策略详解

重试策略（Retry）
// 简单重试3次var retryPolicy = Policy .Handle<Exception>() .Retry(3);// 带退避算法的重试（指数退避）var retryWithBackoffPolicy = Policy .Handle<Exception>() .WaitAndRetryAsync( retryCount: 5, sleepDurationProvider: retryAttempt => TimeSpan.FromSeconds(Math.Pow(2, retryAttempt)), // 2, 4, 8, 16, 32秒 onRetry: (exception, sleepDuration, retryCount, context) => { Console.WriteLine($\"重试 {retryCount}: 等待 {sleepDuration.TotalSeconds}s 后重试，异常: {exception.Message}\"); });
断路器策略（Circuit Breaker）
// 简单断路器：连续3次失败后断开，断开30秒var circuitBreakerPolicy = Policy .Handle<Exception>() .CircuitBreaker( exceptionsAllowedBeforeBreaking: 3, durationOfBreak: TimeSpan.FromSeconds(30), onBreak: (ex, breakDelay) => Console.WriteLine($\"断路器断开: {ex.Message}\"), onReset: () => Console.WriteLine(\"断路器重置\"), onHalfOpen: () => Console.WriteLine(\"断路器半开状态\") );// 高级断路器：基于失败比例（统计窗口内）var advancedCircuitBreaker = Policy .Handle<Exception>() .AdvancedCircuitBreaker( failureThreshold: 0.5, // 50%的请求失败 samplingDuration: TimeSpan.FromSeconds(30), // 统计30秒内的请求 minimumThroughput: 10, // 至少10次请求才开始统计 durationOfBreak: TimeSpan.FromSeconds(30) );
超时策略（Timeout）
// 超时策略：操作超过5秒则终止var timeoutPolicy = Policy .Timeout(TimeSpan.FromSeconds(5), TimeoutStrategy.Pessimistic);// 使用悲观超时策略并处理超时异常try{ await timeoutPolicy.ExecuteAsync(async () => { // 可能耗时较长的操作 await Task.Delay(10000); return \"操作完成\"; });}catch (TimeoutRejectedException ex){ Console.WriteLine($\"操作超时: {ex.Message}\");}
隔离舱策略（Bulkhead Isolation）
// 隔离舱：限制并发执行的操作数量var bulkheadPolicy = Policy .Bulkhead( maxParallelization: 10, // 最多10个并发操作 maxQueuingActions: 5, // 最多5个操作排队等待 onBulkheadRejected: context => Console.WriteLine(\"隔离舱已满，请求被拒绝\") );// 使用隔离舱策略await bulkheadPolicy.ExecuteAsync(async () =>{ // 受隔离舱保护的操作 await DoWorkAsync();});
降级策略（Fallback）
// 降级策略：操作失败时提供备选值var fallbackPolicy = Policy .Handle<Exception>() .FallbackAsync( fallbackValue: \"默认值\", fallbackAction: async ct => // 降级操作 { return await CacheService.GetLatestDataAsync(); }, onFallbackAsync: ex => { Console.WriteLine($\"执行失败，使用降级策略: {ex.Exception.Message}\"); return Task.CompletedTask; } );// 使用降级策略var result = await fallbackPolicy.ExecuteAsync(async () =>{ // 可能失败的操作 return await GetDataFromServiceAsync();});
缓存策略 (Cache)
// 内存缓存策略var cachePolicy = Policy.Cache( new MemoryCacheProvider(myMemoryCache), TimeSpan.FromMinutes(5)); // TTL=5分钟// 使用示例var result = cachePolicy.Execute(context => ExpensiveOperation(), new Context(\"cache_key\"));
策略包装（Policy Wrap）
var policyWrap = Policy.WrapAsync(fallbackPolicy, retryPolicy, circuitBreakerPolicy, timeoutPolicy);var result = await policyWrap.ExecuteAsync(ct => _context.Users.FirstOrDefaultAsync(ct), cancellationToken);
健康检查策略
// 定义健康检查策略var healthPolicy = Policy .Handle<Exception>() .AdvancedCircuitBreaker(...) .WrapAsync(Policy.TimeoutAsync(TimeSpan.FromSeconds(3)));// 健康检查端点app.MapGet(\"/health\", async () => { var healthResult = await healthPolicy.ExecuteAsync(async () => { await db.CheckHealthAsync(); await queue.CheckHealthAsync(); }); return healthResult.IsHealthy ? Results.Ok() : Results.StatusCode(503);});

组合策略

Polly 允许将多个策略组合成一个策略链（Policy Wrap）：

// 组合重试、断路器和超时策略var policyWrap = Policy.WrapAsync( circuitBreakerPolicy, // 最外层：断路器 retryWithBackoffPolicy, // 中间层：带退避的重试 timeoutPolicy // 最内层：超时);// 使用组合策略执行操作await policyWrap.ExecuteAsync(async () =>{ // 受多重策略保护的操作 var response = await httpClient.GetAsync(\"https://api.example.com/data\"); response.EnsureSuccessStatusCode(); return await response.Content.ReadAsStringAsync();});

策略上下文（Context）

通过 Context 可以在策略执行过程中传递自定义数据：

// 创建带上下文的策略var policy = Policy .Handle<Exception>() .RetryAsync(3, (ex, retryCount, context) => { // 从上下文中获取数据 if (context.ContainsKey(\"RequestId\")) { Console.WriteLine($\"请求 {context[\"RequestId\"]} 重试 {retryCount}: {ex.Message}\"); } });// 执行策略时传递上下文await policy.ExecuteAsync( async context => { // 执行受保护的操作 return await DoWorkAsync(); }, new Context { { \"RequestId\", Guid.NewGuid().ToString() } });

Polly 与 HttpClient 集成

使用 Polly 与 HttpClientFactory
// 注册 HttpClient 并应用 Polly 策略services.AddHttpClient(\"ResilientClient\") .AddPolicyHandler(GetRetryPolicy()) .AddPolicyHandler(GetCircuitBreakerPolicy());// 定义重试策略private IAsyncPolicy<HttpResponseMessage> GetRetryPolicy(){ return HttpPolicyExtensions .HandleTransientHttpError() // 处理常见HTTP错误（408, 500-599） .OrResult(msg => msg.StatusCode == System.Net.HttpStatusCode.NotFound) .WaitAndRetryAsync( retryCount: 3, sleepDurationProvider: retryAttempt => TimeSpan.FromSeconds(Math.Pow(2, retryAttempt)), onRetry: (response, sleepDuration, retryCount, context) => { Console.WriteLine($\"HTTP请求失败，重试 {retryCount}: {response.Result?.StatusCode}\"); } );}// 定义断路器策略private IAsyncPolicy<HttpResponseMessage> GetCircuitBreakerPolicy(){ return HttpPolicyExtensions .HandleTransientHttpError() .CircuitBreakerAsync( exceptionsAllowedBeforeBreaking: 3, durationOfBreak: TimeSpan.FromSeconds(30) );}
在控制器中使用
public class MyController : ControllerBase{ private readonly HttpClient _httpClient; public MyController(IHttpClientFactory clientFactory) { _httpClient = clientFactory.CreateClient(\"ResilientClient\"); } [HttpGet] public async Task<IActionResult> GetData() { var response = await _httpClient.GetAsync(\"https://api.example.com/data\"); response.EnsureSuccessStatusCode(); return Ok(await response.Content.ReadAsStringAsync()); }}

高级特性

策略注册表（Policy Registry）
// 注册策略到注册表var policyRegistry = new PolicyRegistry{ { \"RetryPolicy\", Policy.Handle<Exception>().Retry(3) }, { \"CircuitBreakerPolicy\", Policy.Handle<Exception>().CircuitBreaker(3, TimeSpan.FromSeconds(30)) }};// 从注册表获取策略var retryPolicy = policyRegistry.Get<IAsyncPolicy>(\"RetryPolicy\");
自定义策略
// 创建自定义策略：记录所有操作var loggingPolicy = Policy .Handle<Exception>() .ExecuteAndCaptureAsync(async (ctx, cancellationToken) => { Console.WriteLine($\"开始执行操作: {ctx.OperationKey}\"); try { var result = await ctx.CancellationToken.HasValue ? ctx.Action(ctx, ctx.CancellationToken.Value) : ctx.Action(ctx, CancellationToken.None); Console.WriteLine($\"操作成功: {ctx.OperationKey}\"); return result; } catch (Exception ex) { Console.WriteLine($\"操作失败: {ctx.OperationKey}, 异常: {ex.Message}\"); throw; } });
分布式断路器
// 使用Redis实现分布式断路器public class RedisCircuitBreaker : AsyncCircuitBreakerPolicy{ private readonly IDistributedCache _cache; public RedisCircuitBreaker(IDistributedCache cache) { _cache = cache; } protected override async Task<bool> IsolateCircuitAsync(...) { var state = new CircuitState(State.Closed); await _cache.SetAsync(\"CircuitState\", state); } protected override async Task<CircuitState> GetCircuitStateAsync() { return await _cache.GetAsync<CircuitState>(\"CircuitState\"); }}
自适应弹性策略
// 基于响应时间自动调整超时var adaptiveTimeout = Policy.TimeoutAsync(() => { var avgResponseTime = Telemetry.GetAvgResponseTime(\"ServiceA\"); return avgResponseTime * 1.5; // 平均响应时间的1.5倍});
gRPC 弹性处理
// gRPC 弹性策略var grpcPolicy = Policy .Handle<RpcException>(ex => ex.StatusCode == StatusCode.Unavailable || ex.StatusCode == StatusCode.DeadlineExceeded) .WaitAndRetryAsync(3, attempt => TimeSpan.FromSeconds(attempt * 2));// 使用var client = new Greeter.GreeterClient(channel);var response = await grpcPolicy.ExecuteAsync(() => client.SayHelloAsync(new HelloRequest()).ResponseAsync);

性能优化策略

策略重用与缓存
// 策略工厂模式（避免重复创建）public static class PolicyFactory{ private static ConcurrentDictionary<string, IAsyncPolicy> _policies = new(); public static IAsyncPolicy GetOrCreatePolicy(string key, Func<IAsyncPolicy> createPolicy) { return _policies.GetOrAdd(key, _ => createPolicy()); }}// 使用示例var policy = PolicyFactory.GetOrCreatePolicy(\"CircuitBreakerPolicy\", () => Policy.Handle<Exception>() .CircuitBreakerAsync(5, TimeSpan.FromSeconds(30));
上下文传递与监控
// 策略上下文配置var context = new Context{ {\"OperationName\", \"GetUserData\"}, {\"RequestId\", Guid.NewGuid()}};// 策略执行await policy.ExecuteAsync(async ctx => { var requestId = ctx[\"RequestId\"]; Telemetry.StartTracking(requestId); // 业务逻辑}, context);// 事件钩子var policyWithEvents = Policy .Handle<Exception>() .CircuitBreakerAsync(5, TimeSpan.FromSeconds(30), onBreak: (ex, breakDelay, ctx) => { Telemetry.TrackCircuitBreak(ctx[\"OperationName\"].ToString()); }, onReset: ctx => { Telemetry.TrackCircuitReset(ctx[\"OperationName\"].ToString()); });
错误处理模式
try{ return await _resilientPolicy.ExecuteAsync(BusinessOperation);}catch (BrokenCircuitException){ // 处理开路状态 return ServiceUnavailable();}catch (TimeoutRejectedException){ // 处理超时 return GatewayTimeout();}catch (BulkheadRejectedException){ // 处理限流 return TooManyRequests();}catch (Exception ex){ // 其他异常 return InternalError(ex);}
Polly架构黄金法则：
Policy.WrapAsync( FallbackPolicy.ForDegradedService(), // 降级保护 TimeoutPolicy.ForSlowServices(), // 超时控制 CircuitBreakerPolicy.ForProtection(), // 短路保护 BulkheadPolicy.ForConcurrency(), // 资源保护 RetryPolicy.ForTransientFaults() // 重试机制);

混沌工程与性能优化

V8 版本引入 混沌工程策略，模拟生产环境故障以验证系统韧性

builder.AddResilienceHandler(\"chaos-pipeline\", pipelineBuilder => { pipelineBuilder .AddRetry(new RetryStrategyOptions { /*...*/ }) .AddChaosFault(0.02, () => new InvalidOperationException(\"模拟故障\")) // 2% 请求注入异常 .AddChaosLatency(0.01, TimeSpan.FromSeconds(5)); // 1% 请求增加 5s 延迟});

  • 混沌注入类型：故障(Fault)、错误结果(Outcome)、延迟(Latency)、自定义行为(Behavior)

  • 动态控制：通过 InjectionRate 和 EnabledGenerator 动态调整注入频率

集成应用实践

与 HttpClient 深度集成

通过 Microsoft.Extensions.Http.Polly 包实现声明式策略配置

services.AddHttpClient(\"ResilientClient\") .AddPolicyHandler(Policy.TimeoutAsync<HttpResponseMessage>(10)) .AddTransientHttpErrorPolicy(policyBuilder => policyBuilder.WaitAndRetryAsync(3, attempt => TimeSpan.FromSeconds(attempt)));
ASP.NET Core 全局策略注册
services.AddSingleton<AsyncRetryPolicy>(Policy .Handle<SqlException>() .WaitAndRetryAsync(3, attempt => TimeSpan.FromSeconds(2)));services.AddMvc().AddPollyPolicies(); // 配合 AOP 框架如 AspectCore 使用

高级应用场景

电商订单系统熔断设计

  • 触发条件：支付服务连续 5 次超时或返回 5xx 错误

  • 熔断动作：开启 30 秒熔断，快速返回“服务暂不可用”提示

  • 半开状态：熔断结束后首个请求成功则闭合，否则再次熔断

金融交易系统隔板隔离

  • 资源配置：核心交易线程池（10 线程），报表查询线程池（20 线程）

  • 效果：交易高负载时，报表查询不受影响，避免级联故障

性能优化

  • 策略复用：单例模式管理 Policy 对象，避免重复创建

  • 异步支持：优先使用 RetryAsync、ExecuteAsync 异步方法

  • 上下文传递：利用 Context 对象传递参数，支持动态策略调整

混沌工程

混沌工程是通过在生产或类生产环境中故意引入故障（如异常、延迟、假结果），测试系统在异常条件下的行为，以增强系统对不可预测故障的弹性（Resilience）。它源自 Netflix 的 Simian Army 项目（如 Chaos Monkey），旨在验证系统是否能承受“混乱”场景。其核心目标是：

  • 暴露系统脆弱点：模拟生产环境中的非常规故障（如网络抖动、服务不可用）。

  • 检验弹性策略有效性：验证重试、熔断等策略是否能正确处理注入的故障。

  • 提升系统韧性：通过可控的混沌实验，增强系统抵御真实故障的能力。

Polly V8 的优势：

  • 无缝集成：与 Polly 的弹性策略（如重试、断路器）结合，测试弹性机制。

  • 灵活配置：支持同步/异步操作，统一 API（ChaosStrategyOptions）

  • 生产环境测试：在生产环境中以可控方式注入故障，无需修改核心代码。

  • telemetry 支持：通过 Polly 的遥测（Telemetry）系统记录混沌事件。

常见混沌实验类型：

  • 服务响应超时

  • 随机抛出异常

  • 模拟网络抖动

  • 限制资源（CPU、内存）

使用场景

  • HTTP 请求：测试 API 调用在网络延迟或失败时的弹性。

  • 数据库操作：模拟 EF Core 的 SqlException 或 DbUpdateConcurrencyException。

  • 后台任务：结合 BackgroundService 和 ConcurrentQueue 测试批量任务的容错。

  • CQRS 架构：在命令/查询处理程序中注入故障，验证业务逻辑的鲁棒性。

  • 微服务：模拟服务间通信失败，测试分布式系统的可靠性。

核心混沌策略

Polly V8 提供四类策略，支持细粒度故障模拟：

策略类型 作用 典型场景 故障注入 (Fault) 抛出自定义异常 模拟服务崩溃、依赖失效 结果注入 (Outcome) 返回错误结果或异常 测试客户端对错误响应的处理逻辑 延迟注入 (Latency) 增加处理延迟 模拟网络拥塞或资源竞争 行为注入 (Behavior) 执行自定义操作（如重启服务） 验证系统自愈能力
精细化控制参数
参数 类型 默认值 作用 InjectionRate double (0-1) 0.001 混沌触发概率（如0.05表示5%的请求受影响） InjectionRateGenerator Func null 动态生成注入率（支持按请求上下文调整） EnabledGenerator Func null 动态启用策略（例如仅在测试环境激活）
核心混沌策略
Fault（故障注入）

注入异常（如 TimeoutException），模拟系统错误。

var faultOptions = new ChaosFaultStrategyOptions{ FaultGenerator = new FaultGenerator() .AddException<TimeoutException>() .AddException(() => new InvalidOperationException(\"Chaos fault\")), InjectionRate = 0.1, // 10% 概率 OnFaultInjected = args => { Console.WriteLine($\"Fault injected: {args.Fault.Message}\"); return default; }};new ResiliencePipelineBuilder().AddChaosFault(faultOptions);
Outcome（结果注入）

注入假结果或异常，测试应用对不同响应的处理。

var outcomeOptions = new ChaosOutcomeStrategyOptions{ OutcomeGenerator = args => new ValueTask<Outcome<object>>(new Outcome<object>(42)), // 注入假结果 42 InjectionRate = 0.1, OnOutcomeInjected = args => { Console.WriteLine($\"Outcome injected: {args.Outcome.Result}\"); return default; }};new ResiliencePipelineBuilder().AddChaosOutcome(outcomeOptions);
Latency（延迟注入）

在操作前引入延迟，模拟网络或服务瓶颈。

var latencyOptions = new ChaosLatencyStrategyOptions{ Latency = TimeSpan.FromSeconds(5), InjectionRate = 0.05, OnLatencyInjected = args => { Console.WriteLine($\"Latency injected: {args.Latency.TotalSeconds} seconds\"); return default; }};new ResiliencePipelineBuilder().AddChaosLatency(latencyOptions);
Behavior（行为注入）

在操作前执行自定义行为（如重启 Redis）。

var behaviorOptions = new ChaosBehaviorStrategyOptions{ BehaviorGenerator = args => RestartRedisAsync(args.Context.CancellationToken), InjectionRate = 0.05, OnBehaviorInjected = args => { Console.WriteLine(\"Behavior injected\"); return default; }};new ResiliencePipelineBuilder().AddChaosBehavior(behaviorOptions);async Task RestartRedisAsync(CancellationToken ct){ // 模拟重启 Redis await Task.Delay(100, ct);}
组合混沌策略
var pipeline = new ResiliencePipelineBuilder() .AddRetry(new RetryStrategyOptions { MaxRetryAttempts = 3, Delay = TimeSpan.FromSeconds(2) }) .AddChaosFault(new ChaosFaultStrategyOptions { FaultGenerator = new FaultGenerator().AddException<TimeoutException>(), InjectionRate = 0.1 }) .AddChaosLatency(new ChaosLatencyStrategyOptions { Latency = TimeSpan.FromSeconds(5), InjectionRate = 0.05 }) .Build();
// 组合异常注入和延迟注入策略var chaosPolicy = Policy.WrapAsync( faultInjectionPolicy, latencyInjectionPolicy);// 执行受混沌策略影响的操作await chaosPolicy.ExecuteAsync(async () =>{ // 可能会被注入故障或延迟的操作 return await FetchDataFromServiceAsync();});
混沌实验设计与执行
实验参数配置
// 使用配置对象创建混沌策略var chaosSettings = new ChaosSettings{ Enabled = true, FaultType = FaultType.Exception, // 或 Latency InjectionRate = 0.1, // 10%的故障率 ExceptionType = typeof(TransientFailureException), Latency = TimeSpan.FromSeconds(1)};var chaosPolicy = Policy .HandleResult<string>(_ => true) .InjectChaosAsync(chaosSettings);
渐进式混沌实验
// 随着时间增加故障率var injectionRateProvider = new InjectionRateSlidingWindow( initialInjectionRate: 0.05, // 初始5%故障率 maxInjectionRate: 0.2, // 最大20%故障率 windowDuration: TimeSpan.FromMinutes(5) // 每5分钟增加一次);var chaosPolicy = Policy .HandleResult<int>(_ => true) .InjectExceptionAsync( injectionRateProvider: () => injectionRateProvider.CurrentInjectionRate, faultFactory: () => new TransientFailureException(\"系统压力测试\") );
与弹性策略的管道集成

混沌策略需置于标准弹性策略之后，形成“先防御后攻击”的管道

services.AddHttpClient(\"ResilientClient\") .AddResilienceHandler(\"chaos-pipeline\", builder => { // 标准弹性策略 builder .AddRetry(new RetryStrategyOptions { MaxRetryAttempts = 3 }) .AddCircuitBreaker(new CircuitBreakerStrategyOptions { FailureRatio = 0.3 }) .AddTimeout(TimeSpan.FromSeconds(10)); // 混沌策略（置于弹性策略后） builder .AddChaosLatency(0.05, TimeSpan.FromSeconds(15)) .AddChaosFault(0.03, () => new HttpRequestException(\"Chaos fault\")); });// 创建弹性策略（重试 + 断路器）var resiliencePolicy = Policy.WrapAsync( Policy.Handle<Exception>().RetryAsync(3), Policy.Handle<Exception>().CircuitBreakerAsync(3, TimeSpan.FromSeconds(30)));// 创建混沌策略var chaosPolicy = Policy .HandleResult<string>(_ => true) .InjectExceptionAsync(0.1, () => new HttpRequestException(\"模拟网络故障\"));// 组合策略：先应用混沌，再应用弹性var combinedPolicy = Policy.WrapAsync(resiliencePolicy, chaosPolicy);// 执行受保护的操作await combinedPolicy.ExecuteAsync(async () =>{ return await httpClient.GetStringAsync(\"https://api.example.com/data\");});

管道执行顺序：
请求 → 重试 → 熔断 → 超时 → 注入延迟/故障 → 目标服务

在测试环境中使用
单元测试示例
[Fact]public async Task GetData_ShouldSucceed_WhenTransientFailureOccurs(){ // 安排：创建带有故障注入的策略 var chaosPolicy = Policy .HandleResult<string>(_ => true) .InjectExceptionAsync( injectionRate: 0.5, // 50%的故障率 faultFactory: () => new HttpRequestException(\"模拟网络故障\") ); // 安排：创建带有重试策略的服务 var service = new MyService( httpClient: _httpClient, resiliencePolicy: Policy.WrapAsync( Policy.Handle<HttpRequestException>().RetryAsync(3), chaosPolicy ) ); // 执行 var result = await service.GetDataAsync(); // 断言：尽管有故障注入，服务仍应成功返回数据 Assert.NotNull(result);}
集成测试配置
// 在测试环境中启用混沌工程public void ConfigureServices(IServiceCollection services){ if (Environment.IsDevelopment() || Environment.IsStaging()) { // 添加混沌策略 services.AddTransient<MyService>(sp => { var httpClient = sp.GetRequiredService<HttpClient>(); // 创建混沌策略 var chaosPolicy = Policy .HandleResult<string>(_ => true) .InjectChaosAsync(new ChaosSettings {  Enabled = true,  FaultType = FaultType.Exception,  InjectionRate = 0.1,  ExceptionType = typeof(HttpRequestException) }); // 创建弹性策略 var resiliencePolicy = Policy.WrapAsync( Policy.Handle<Exception>().RetryAsync(3), chaosPolicy ); return new MyService(httpClient, resiliencePolicy); }); } else { // 在生产环境中不启用混沌 services.AddTransient<MyService>(); }}

ASP.NET Core Web 项目中 Polly 实战

基础环境配置
  • 安装 NuGet 包
dotnet add package Microsoft.Extensions.Http.Pollydotnet add package Pollydotnet add package Polly.Contrib.WaitAndRetry # 提供高级重试策略
入口文件配置
// Program.csbuilder.Services.AddHttpClient(\"ResilientApi\") .AddTransientHttpErrorPolicy(policyBuilder => policyBuilder.WaitAndRetryAsync( Backoff.DecorrelatedJitterBackoffV2( medianFirstRetryDelay: TimeSpan.FromSeconds(1),  retryCount: 3 ) ) );
核心应用场景实现
外部 API 调用保护（商品服务）
// ProductService.cspublic class ProductService{ private readonly IHttpClientFactory _httpClientFactory; public ProductService(IHttpClientFactory httpClientFactory) { _httpClientFactory = httpClientFactory; } public async Task<Product> GetProductAsync(int id) { var client = _httpClientFactory.CreateClient(\"ResilientApi\"); // 带策略的执行 return await client.GetFromJsonAsync<Product>($\"/api/products/{id}\"); }}// 策略说明：// 1. 自动处理5xx错误、408超时和网络故障// 2. 使用指数退避算法重试3次：0.5s, 1.5s, 4.5s// 3. 添加Jitter避免同步重试风暴
数据库操作保护（EF Core 集成）
// OrderRepository.cspublic class OrderRepository{ private readonly AsyncRetryPolicy _dbPolicy; private readonly AppDbContext _context; public OrderRepository(AppDbContext context) { _context = context; // 配置SQL重试策略 _dbPolicy = Policy .Handle<SqlException>(ex => ex.IsTransient) // 捕获瞬态错误 .Or<TimeoutException>() .WaitAndRetryAsync( retryCount: 3, sleepDurationProvider: attempt => TimeSpan.FromSeconds(Math.Pow(2, attempt)), onRetry: (ex, delay) =>  Log.Warning($\"数据库操作失败，{delay.TotalSeconds}秒后重试: {ex.Message}\") ); } public async Task SaveOrderAsync(Order order) { await _dbPolicy.ExecuteAsync(async () => { _context.Orders.Add(order); await _context.SaveChangesAsync(); }); }}
添加降级策略

如果外部服务不可用，可以添加降级策略提供默认响应：

// 在Program.cs中添加降级策略var fallbackPolicy = Policy<HttpResponseMessage> .Handle<Exception>() .FallbackAsync( fallbackAction: async (ctx, token) => { Console.WriteLine(\"执行降级策略，返回缓存数据\"); var response = new HttpResponseMessage(HttpStatusCode.OK); response.Content = new StringContent(\"[{\\\"Temperature\\\":25,\\\"Condition\\\":\\\"Sunny\\\"}]\"); return response; }, onFallbackAsync: ex => { Console.WriteLine($\"请求失败，执行降级: {ex.Exception.Message}\"); return Task.CompletedTask; } );// 更新HttpClient注册，添加降级策略builder.Services.AddHttpClient(\"ResilientClient\") .AddPolicyHandler(fallbackPolicy) .AddPolicyHandler(retryPolicy) .AddPolicyHandler(circuitBreakerPolicy);
全局熔断保护（防止服务雪崩）
// Program.csvar circuitBreakerPolicy = Policy .Handle<HttpRequestException>() .Or<TimeoutException>() .CircuitBreakerAsync( exceptionsAllowedBeforeBreaking: 5, durationOfBreak: TimeSpan.FromSeconds(30), onBreak: (ex, breakDelay) => Log.Error($\"熔断器开启！{breakDelay.TotalSeconds}秒内拒绝请求\"), onReset: () => Log.Information(\"熔断器关闭\") );builder.Services.AddHttpClient(\"PaymentService\") .AddPolicyHandler(circuitBreakerPolicy) .AddPolicyHandler(Policy.TimeoutAsync<HttpResponseMessage>(10));
高级策略组合
完整策略链（重试+熔断+超时）
// Program.csvar retryPolicy = Policy<HttpResponseMessage> .HandleResult(r => (int)r.StatusCode >= 500) .Or<HttpRequestException>() .WaitAndRetryAsync(3, attempt => TimeSpan.FromSeconds(0.5 * Math.Pow(2, attempt)));var circuitBreaker = Policy<HttpResponseMessage> .HandleResult(r => (int)r.StatusCode >= 500) .CircuitBreakerAsync(5, TimeSpan.FromSeconds(30));var timeoutPolicy = Policy.TimeoutAsync<HttpResponseMessage>(10);// 策略组合顺序：超时 → 熔断 → 重试var policyWrap = Policy.WrapAsync(timeoutPolicy, circuitBreaker, retryPolicy);builder.Services.AddHttpClient(\"InventoryService\") .AddPolicyHandler(policyWrap);
环境差异化配置
// 开发环境更激进的策略if (env.IsDevelopment()){ services.AddHttpClient(\"DevService\") .AddTransientHttpErrorPolicy(builder => builder.WaitAndRetryAsync(5, attempt => TimeSpan.FromSeconds(0.5)));}// 生产环境保守策略else { services.AddHttpClient(\"ProdService\") .AddPolicyHandler(Policy.TimeoutAsync<HttpResponseMessage>(3)) .AddTransientHttpErrorPolicy(builder => builder.CircuitBreakerAsync(5, TimeSpan.FromSeconds(30)));}
监控与日志集成
// Serilog 集成Log.Logger = new LoggerConfiguration() .Enrich.WithProperty(\"Application\", \"ECommerce\") .WriteTo.Console() .WriteTo.Seq(\"http://localhost:5341\") .CreateLogger();// Polly 事件记录policy = policy.WithPolicyLogging(logger, \"InventoryPolicy\");
策略事件追踪
var policy = Policy .Handle<Exception>() .RetryAsync(3, onRetry: (ex, retryCount) => { logger.LogWarning(\"重试 #{RetryCount} 触发: {Exception}\", retryCount, ex); }) .WithTelemetry(loggerFactory); // 启用Polly内置遥测
模拟故障测试
// 测试环境注入故障if (env.IsEnvironment(\"Test\")){ services.AddHttpClient(\"MockService\") .AddPolicyHandler(Policy<HttpResponseMessage> .HandleResult(r => true) .FallbackAsync(new HttpResponseMessage(HttpStatusCode.ServiceUnavailable))) .AddPolicyHandler(new FaultInjectionPolicy(0.1)); // 10%失败率}
配置建议
配置驱动策略参数
// appsettings.json\"PollyConfig\": { \"HttpRetryCount\": 3, \"HttpCircuitBreakThreshold\": 5, \"HttpCircuitBreakDuration\": \"00:00:30\", \"HttpTimeout\": \"00:00:15\", \"DbRetryCount\": 3, \"DbRetryDelayMilliseconds\": 200}
// 基于配置的策略static IAsyncPolicy<HttpResponseMessage> GetConfiguredPolicy(IConfiguration config){ var retryCount = config.GetValue<int>(\"PollyConfig:HttpRetryCount\"); var circuitBreak = config.GetValue<int>(\"PollyConfig:HttpCircuitBreakThreshold\"); var breakDuration = config.GetValue<TimeSpan>(\"PollyConfig:HttpCircuitBreakDuration\"); var timeout = config.GetValue<TimeSpan>(\"PollyConfig:HttpTimeout\"); return Policy.WrapAsync( Policy.TimeoutAsync<HttpResponseMessage>(timeout), HttpPolicyExtensions .HandleTransientHttpError() .CircuitBreakerAsync(circuitBreak, breakDuration), HttpPolicyExtensions .HandleTransientHttpError() .WaitAndRetryAsync(retryCount, retryAttempt =>  TimeSpan.FromSeconds(Math.Pow(2, retryAttempt))) );}
Web 项目特定场景处理
控制器动作级保护
[HttpGet(\"products/{id}\")]public async Task<IActionResult> GetProduct(int id){ // 创建请求级策略 var policy = Policy .Handle<Exception>() .RetryAsync(2, (ex, retryCount) => Log.Warning($\"第{retryCount}次重试: {ex.Message}\")); var product = await policy.ExecuteAsync( () => _productService.GetProductAsync(id)); return product != null ? Ok(product) : NotFound();}
自定义响应转换
// 处理外部API的429响应var rateLimitPolicy = Policy<HttpResponseMessage> .HandleResult(r => r.StatusCode == HttpStatusCode.TooManyRequests) .RetryAsync(1, onRetry: (result, retryCount) => { var retryAfter = result.Result.Headers.RetryAfter?.Delta ?? TimeSpan.FromSeconds(5); Thread.Sleep(retryAfter.Value); });builder.Services.AddHttpClient(\"ExternalApi\") .AddPolicyHandler(rateLimitPolicy);
用户请求上下文传递
var policy = Policy .Handle<Exception>() .RetryAsync(1, onRetry: (ex, retryCount, context) => { var userId = context[\"user-id\"]?.ToString(); Log.Warning($\"用户{userId}请求重试: {ex.Message}\"); });// 执行时传递上下文var response = await policy.ExecuteAsync(ctx => _service.CallApi(userId), new Context { [\"user-id\"] = userId });
下单流程示例
// OrderController.cs[HttpPost(\"orders\")]public async Task<IActionResult> CreateOrder([FromBody] OrderRequest request){ // 组合策略: 超时5s + 3次重试 var policy = Policy .Handle<TimeoutException>() .Or<HttpRequestException>() .WaitAndRetryAsync(3, attempt => TimeSpan.FromSeconds(attempt)) .WrapAsync(Policy.TimeoutAsync(5)); try { var result = await policy.ExecuteAsync(async () => { // 1. 检查库存 var stockValid = await _inventoryService.CheckStockAsync(request.Items); // 2. 创建支付 var payment = await _paymentService.CreatePaymentAsync(request.Payment); // 3. 保存订单 var order = await _orderService.CreateOrderAsync(request, payment.Id); return new { OrderId = order.Id }; }); return Ok(result); } catch (TimeoutException) { return StatusCode(503, \"服务响应超时，请稍后重试\"); } catch (HttpRequestException) { return StatusCode(502, \"上游服务不可用\"); }}
网络标签：

相关问题