From cb4ecd39515784ea3e5db17e43a9300e8e992f76 Mon Sep 17 00:00:00 2001 From: Jason Date: Wed, 13 May 2026 23:20:45 +0800 Subject: [PATCH] fix(proxy): refine failover decisions in forwarder MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Two related changes to make per-provider failover behave correctly. 1. Bucket UpstreamError by status code in categorize_proxy_error. The old "every UpstreamError is Retryable" rule meant a malformed client request (400 / 422) would be replayed against every provider in the queue: errors amplified N-fold, the circuit breaker accrued unwarranted failure counts, and quota was burned. Now 400 / 405 / 406 / 413 / 414 / 415 / 422 / 501 are NonRetryable since the request itself is wrong and no provider will accept it. 401 / 403 / 404 / 408 / 409 / 429 / 451 and all 5xx remain Retryable because the next provider may carry a different key, quota, region, or model mapping. 2. Make the rectifier-retry path participate in failover. Both the signature (RECT-003) and budget (RECT-012) rectifier branches used to "return Err(...)" after the retry failed, short-circuiting the per-provider loop. A provider-side failure (5xx / Timeout / ForwardFailed) now records the circuit breaker, accumulates into last_error / last_provider, and "continue"s to the next provider — matching the normal Retryable arm. Client-side failures still return immediately since a different provider cannot fix a malformed payload. --- src-tauri/src/proxy/forwarder.rs | 82 ++++++++++++++++++++++---------- 1 file changed, 58 insertions(+), 24 deletions(-) diff --git a/src-tauri/src/proxy/forwarder.rs b/src-tauri/src/proxy/forwarder.rs index bef812436..230b617db 100644 --- a/src-tauri/src/proxy/forwarder.rs +++ b/src-tauri/src/proxy/forwarder.rs @@ -422,12 +422,14 @@ impl RequestForwarder { }); } Err(retry_err) => { - // 整流重试仍失败：区分错误类型决定是否记录熔断器 + // 整流重试仍失败：区分错误类型 log::warn!( "[{app_type_str}] [RECT-003] 整流重试仍失败: {retry_err}" ); - // 区分错误类型：Provider 问题记录失败，客户端问题仅释放 permit + // Provider 错误 vs 客户端错误：Provider 错误说明本家 + // 上游有问题，下一家 provider 可能可用，应当继续故障转移； + // 客户端错误说明整流后的请求本身就有问题，下一家也修不好。 let is_provider_error = match &retry_err { ProxyError::Timeout(_) | ProxyError::ForwardFailed(_) => true, @@ -438,7 +440,7 @@ impl RequestForwarder { }; if is_provider_error { - // Provider 问题：记录失败到熔断器 + // Provider 问题：记录熔断器后继续轮询下一家 let _ = self .router .record_result( @@ -449,17 +451,26 @@ impl RequestForwarder { Some(retry_err.to_string()), ) .await; - } else { - // 客户端问题：仅释放 permit，不记录熔断器 - self.router - .release_permit_neutral( - &provider.id, - app_type_str, - used_half_open_permit, - ) - .await; + { + let mut status = self.status.write().await; + status.last_error = Some(format!( + "Provider {} 整流重试失败: {}", + provider.name, retry_err + )); + } + last_error = Some(retry_err); + last_provider = Some(provider.clone()); + continue; } + // 客户端问题：释放 permit 后直接返回，不再尝试其它 provider + self.router + .release_permit_neutral( + &provider.id, + app_type_str, + used_half_open_permit, + ) + .await; let mut status = self.status.write().await; status.failed_requests += 1; status.last_error = Some(retry_err.to_string()); @@ -615,6 +626,8 @@ impl RequestForwarder { "[{app_type_str}] [RECT-012] budget 整流重试仍失败: {retry_err}" ); + // 与 signature 整流分支保持一致：provider 错误继续故障转移， + // 客户端错误直接返回。 let is_provider_error = match &retry_err { ProxyError::Timeout(_) | ProxyError::ForwardFailed(_) => { true @@ -634,16 +647,25 @@ impl RequestForwarder { Some(retry_err.to_string()), ) .await; - } else { - self.router - .release_permit_neutral( - &provider.id, - app_type_str, - used_half_open_permit, - ) - .await; + { + let mut status = self.status.write().await; + status.last_error = Some(format!( + "Provider {} budget 整流重试失败: {}", + provider.name, retry_err + )); + } + last_error = Some(retry_err); + last_provider = Some(provider.clone()); + continue; } + self.router + .release_permit_neutral( + &provider.id, + app_type_str, + used_half_open_permit, + ) + .await; let mut status = self.status.write().await; status.failed_requests += 1; status.last_error = Some(retry_err.to_string()); @@ -1654,10 +1676,22 @@ impl RequestForwarder { ProxyError::Timeout(_) => ErrorCategory::Retryable, ProxyError::ForwardFailed(_) => ErrorCategory::Retryable, ProxyError::ProviderUnhealthy(_) => ErrorCategory::Retryable, - // 上游 HTTP 错误：无论状态码如何，都尝试下一个供应商 - // 原因：不同供应商有不同的限制和认证，一个供应商的 4xx 错误 - // 不代表其他供应商也会失败 - ProxyError::UpstreamError { .. } => ErrorCategory::Retryable, + // 上游 HTTP 错误：按状态码分桶。 + // + // 客户端请求自身有问题的状态码无论换哪个 provider 都会被拒绝， + // 继续轮询只会放大错误率、污染熔断器健康度、浪费配额： + // 400 Bad Request / 422 Unprocessable Entity ← 请求体格式或语义错误 + // 405 Method Not Allowed / 406 Not Acceptable ← 方法或 Accept 错误 + // 413 Payload Too Large / 414 URI Too Long ← 客户端构造超限 + // 415 Unsupported Media Type ← Content-Type 错误 + // 501 Not Implemented ← 上游协议确实不支持 + // + // 其他 4xx（401/403/404/408/409/429/451 等）和全部 5xx 都保留 + // Retryable —— 换一家 provider 可能持有不同的 key、配额、地域或模型映射。 + ProxyError::UpstreamError { status, .. } => match *status { + 400 | 405 | 406 | 413 | 414 | 415 | 422 | 501 => ErrorCategory::NonRetryable, + _ => ErrorCategory::Retryable, + }, // Provider 级配置/转换问题：换一个 Provider 可能就能成功 ProxyError::ConfigError(_) => ErrorCategory::Retryable, ProxyError::TransformError(_) => ErrorCategory::Retryable,