.NET6之MiniAPI(二十四)：用Polly重试

为了保障系统的稳定和安全，在调用三方服务时，可以增加重试和熔断。重试是调用一次失败后再试几试，避免下游服务一次闪断，就把整个链路终止；熔断是为了防止太多的次数的无效访问，导致系统不可知异常。

Polly是独立的重试机制的三方库，这里只说明在使用httpclient时，请求下游api时的重试和熔断。需要引入NuGet包Microsoft.Extensions.Http.Polly。

先看一个简单的重试

using Polly;var builder = WebApplication.CreateBuilder(args);builder.Services    .AddHttpClient("RetryClient", httpclient =>    {        httpclient.BaseAddress = new Uri("http://localhost:5258");    })    .AddTransientHttpErrorPolicy(policyBuilder => policyBuilder.RetryAsync(3));var app = builder.Build();//调用httpclientapp.MapGet("/test", async (IHttpClientFactory httpClientFactory) =>{    try    {        var httpClient = httpClientFactory.CreateClient("RetryClient");        var content = await httpClient.GetStringAsync("other-api");        Console.WriteLine(content);        return "ok";    }    catch (Exception exc)    {        if (!Count.Time.HasValue)        {            Count.Time = DateTime.Now;        }        return $"{exc.Message}    【次数：{Count.I++}】  【{Count.Time.Value.ToString("yyyy-MM-dd HH:mm:ss.fffffff")}】";    }});//被调用的接口，返回状态码500app.MapGet("/other-api", (ILogger<Program> logger) =>{    logger.LogInformation($"失败:{DateTime.Now.ToString("yyyy-MM-dd HH:mm:ss.fffffff")}");    return Results.StatusCode(500);});app.Run();static class Count{    public static int I = 1;    public static DateTime? Time;}

返回的结果：

通过AddTransientHttpErrorPolicy(policyBuilder => policyBuilder.RetryAsync(3));我们让请求进行了3次重试，加上原本的一次，就是4次请求。在红色框时，会发现请求集中在极短的时间内，如果下游服务有故障，可能不会在这么短的时间内自动恢复，更好的做法是：根据重试的次数，来延长（或随机，或自建延时算法）请求的时间，比如：

.AddTransientHttpErrorPolicy(policyBuilder =>        policyBuilder.WaitAndRetryAsync(3, retryNumber =>        {            switch (retryNumber)            {                case 1:                    return TimeSpan.FromMilliseconds(500);                case 2:                    return TimeSpan.FromMilliseconds(1000);                case 3:                    return TimeSpan.FromMilliseconds(1500);                default:                    return TimeSpan.FromMilliseconds(100);            }        }));

这时的结果如下，基本按照我们设置的时间来重试的：

还有几种重试策略，如下：

//一直重试.AddTransientHttpErrorPolicy(policyBuilder => policyBuilder.RetryForeverAsync());//每2秒重试一次.AddTransientHttpErrorPolicy(policyBuilder => policyBuilder.WaitAndRetryForeverAsync(retryNumber =>{    Console.WriteLine(retryNumber);    return TimeSpan.FromSeconds(2);}));//在5秒内4次请求，如果50%失败，就熔断10秒.AddTransientHttpErrorPolicy(policyBuilder => policyBuilder.AdvancedCircuitBreakerAsync(0.5d, TimeSpan.FromSeconds(5), 4, TimeSpan.FromSeconds(10)));

熔断是保护服务的手段，在本例中具体用法如下：

builder.Services    .AddHttpClient("RetryClient", httpclient =>    {        httpclient.BaseAddress = new Uri("http://localhost:5258");    })    .AddTransientHttpErrorPolicy(policyBuilder =>        policyBuilder.WaitAndRetryAsync(3, retryNumber =>        {            switch (retryNumber)            {                case 1:                    return TimeSpan.FromMilliseconds(500);                case 2:                    return TimeSpan.FromMilliseconds(1000);                case 3:                    return TimeSpan.FromMilliseconds(1500);                default:                    return TimeSpan.FromMilliseconds(100);            }        }))    //熔断    .AddTransientHttpErrorPolicy(policyBuilder =>  policyBuilder.CircuitBreakerAsync(6, TimeSpan.FromSeconds(30)));

CircuitBreaker控制如果有6次失败的请求，就暂停30秒，具体提示如下: