Merge pull request #5392 from fengshao1227/fix/grok-405-failover-and-cooldown

fix(grok): 405 纳入 failover 与临时踢号，修复 Grok 会话粘性锁死

backend/internal/service/openai_gateway_grok_405_test.go

@@ -0,0 +1,29 @@
+package service
+
+import (
+	"net/http"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+)
+
+func TestShouldFailoverUpstreamError_405IsFailoverEligible(t *testing.T) {
+	svc := &OpenAIGatewayService{}
+
+	assert.True(t, svc.shouldFailoverUpstreamError(http.StatusMethodNotAllowed),
+		"405 should trigger failover so sticky sessions can escape to healthy accounts")
+}
+
+func TestShouldFailoverUpstreamError_ExistingCodesStillWork(t *testing.T) {
+	svc := &OpenAIGatewayService{}
+
+	failoverCodes := []int{401, 402, 403, 405, 429, 529, 500, 502, 503, 504}
+	for _, code := range failoverCodes {
+		assert.True(t, svc.shouldFailoverUpstreamError(code), "status %d should trigger failover", code)
+	}
+
+	nonFailoverCodes := []int{200, 201, 400, 404, 408, 422}
+	for _, code := range nonFailoverCodes {
+		assert.False(t, svc.shouldFailoverUpstreamError(code), "status %d should NOT trigger failover", code)
+	}
+}

backend/internal/service/openai_gateway_upstream_errors.go

@@ -211,7 +211,7 @@ func isOpenAIContextWindowError(upstreamMsg string, upstreamBody []byte) bool {
 
 func (s *OpenAIGatewayService) shouldFailoverUpstreamError(statusCode int) bool {
 	switch statusCode {
-	case 401, 402, 403, 429, 529:
+	case 401, 402, 403, 405, 429, 529:
 		return true
 	default:
 		return statusCode >= 500