Azure虚拟机创建失败（StatusCode:TooManyRequests）：重试策略问题咨询

• 以10秒间隔重试6次创建Azure虚拟机，仍收到TooManyRequests错误，错误详情如下：

Failed with StatusCode: TooManyRequests  
RequestMessage: Method: PUT, RequestUri: 'https://management.azure.com/subscriptions/aaa-bbb-ccc-ddd/resourceGroups/reqource_grp/providers/Microsoft.Compute/virtualMachines/VM_NAME?api-version=2022-03-01', Version: 1.1, Content: System.Net.Http.StringContent, Headers:
{
  Accept: application/json
  Authorization: Bearer ""
  traceparent: 00-14a9ce9cba751f4cfb217ec0b8991ede-b1b327a3be820f60-00
  Content-Type: application/json; charset=utf-8
  Content-Length: 1494
} 
 Content: System.Net.Http.HttpConnectionResponseContent 
 ReasonPhrase:  
 Headers Cache-Control: no-cache
Pragma: no-cache
Retry-After: 10
x-ms-failure-cause: gateway
x-ms-request-id: 345672-331f-4c6a-becc-3f98a9459a5c
x-ms-correlation-request-id: 34563cf-331f-4c6a-becc-3f98a9459a5c
x-ms-routing-request-id: 20240719T104204Z:675a93cf-331f-4c6a-becc-3f98a9459a5c
Strict-Transport-Security: max-age=31536000; includeSubDomains
X-Content-Type-Options: nosniff
X-Cache: CONFIG_NOCACHE
X-MSEdge-Ref: Ref A: 316D2AEF72A1434BBFA503FB3AFC2936 Ref B: AMS231032609023 Ref C: 2024-07-19T10:42:04Z
Date: Fri, 19 Jul 2024 10:42:03 GMT

• 当前使用的Polly重试策略代码：

TimeSpan timeBetweenRetryForRequestUnitLimit = TimeSpan.FromSeconds(10); 
int maxAttemptsToRetry = 6;

Policy.HandleResult<HttpResponseMessage>(response => !response.IsSuccessStatusCode)
    .WaitAndRetryAsync(maxAttemptsToRetry, retryAttempt => {
        log.LogInformation("Success Code Failure Attempt: {retryCount} Payload: {text}", retryAttempt, text);
        return timeBetweenRetryForRequestUnitLimit;
    });

• 补充：响应中未找到x-ms-ratelimit-remaining-subscription-writes请求头

1. 优先遵循Retry-After头动态调整重试间隔 #

响应中已返回Retry-After: 10，但固定间隔无法适配Azure动态变化的限流规则，后续请求的等待时间可能需要更长。修改策略，从响应头提取官方指定的等待时间：

Policy.HandleResult<HttpResponseMessage>(response => 
        response.StatusCode == System.Net.HttpStatusCode.TooManyRequests)
    .WaitAndRetryAsync(maxAttemptsToRetry, (retryAttempt, response, context) => {
        log.LogInformation("TooManyRequests Attempt: {retryCount}", retryAttempt);
        // 提取Retry-After值，无值则用指数退避兜底
        if (response.Result.Headers.TryGetValues("Retry-After", out var values) && 
            int.TryParse(values.First(), out int retryAfterSeconds))
        {
            return TimeSpan.FromSeconds(retryAfterSeconds);
        }
        return TimeSpan.FromSeconds(10 * retryAttempt);
    });

注意：仅针对TooManyRequests错误重试，避免对参数错误等无法通过重试解决的问题无效重试。

2. 改用指数退避策略 #

固定10秒间隔易持续触发限流，改用指数退避（如第1次10秒、第2次20秒、第3次40秒...），给Azure足够的配额恢复时间：

Policy.HandleResult<HttpResponseMessage>(response => 
        response.StatusCode == System.Net.HttpStatusCode.TooManyRequests)
    .WaitAndRetryAsync(maxAttemptsToRetry, retryAttempt => {
        log.LogInformation("TooManyRequests Attempt: {retryCount}", retryAttempt);
        return TimeSpan.FromSeconds(Math.Pow(2, retryAttempt) * 5); // 起始5秒，每次翻倍
    });

3. 理解x-ms-ratelimit-remaining-subscription-writes头缺失原因 #

该头仅在未触发限流的成功响应中返回，收到TooManyRequests时Azure不会返回剩余配额，属于正常行为。若需提前监控配额，需在成功请求中捕获该头的值，提前调整请求频率。

4. 排查并发请求来源 #

确认当前订阅/资源组下是否有其他自动化操作（如批量创建VM、部署脚本）同时执行，这些操作会消耗订阅级请求配额。如有，需协调执行时间，降低并发量。

5. 调整重试参数 #

若业务允许，可适当增加最大重试次数（如至10次）或设置最大等待时间上限，避免过早放弃请求。

内容的提问来源于stack exchange，提问作者Rihab Kasim