Merge pull request #6372 from clansty/fix/openai-oauth-429-fallback-disabled

fix: 修复关闭 429 默认回避后仍触发 OAuth 运行时冷却

backend/internal/service/openai_account_runtime_block_fastpath.go

@@ -37,9 +37,8 @@ const (
 	openAIOAuth429QuotaReset
 )
 
-// classifyOpenAIOAuth429 区分账号配额耗尽信号与普通瞬时 429。明确窗口达到
-// 100% 时以该窗口为准；没有 100% 标记但包含重置头时，沿用 v179 的兼容语义，
-// 仍视为配额限流信号。
+// classifyOpenAIOAuth429 区分账号配额耗尽信号与普通瞬时 429。只有窗口达到
+// 100% 或响应体明确给出 reset 时间时，才视为配额限流信号。
 func classifyOpenAIOAuth429(headers http.Header, responseBody []byte) (openAIOAuth429Disposition, *time.Time) {
 	if snapshot := ParseCodexRateLimitHeaders(headers); snapshot != nil {
 		if normalized := snapshot.Normalize(); normalized != nil {
@@ -228,13 +227,17 @@ func (s *OpenAIGatewayService) markOpenAIOAuth429RateLimited(ctx context.Context
 		return
 	}
 
-	cooldownUntil := time.Now().Add(openAIOAuth429FallbackCooldown)
-	if resetAt != nil && resetAt.After(time.Now()) {
+	now := time.Now()
+	cooldownUntil := now.Add(openAIOAuth429FallbackCooldown)
+	if resetAt != nil && resetAt.After(now) {
 		cooldownUntil = *resetAt
 	} else if s.rateLimitService != nil {
-		if cooldown, ok := s.rateLimitService.get429FallbackCooldown(ctx, account); ok && cooldown > 0 {
-			cooldownUntil = time.Now().Add(cooldown)
+		cooldown, ok := s.rateLimitService.get429FallbackCooldown(ctx, account)
+		if !ok || cooldown <= 0 {
+			s.openaiOAuth429RetryStartedAt.Delete(account.ID)
+			return
 		}
+		cooldownUntil = now.Add(cooldown)
 	}
 	s.BlockAccountScheduling(account, cooldownUntil, "429")
 	s.openaiOAuth429RetryStartedAt.Delete(account.ID)

backend/internal/service/openai_account_runtime_block_fastpath_test.go

@@ -72,6 +72,47 @@ func TestOpenAI429FastPath_BlocksOAuthOnlyAfterRetryWindow(t *testing.T) {
 	require.False(t, svc.shouldRetryOpenAIOAuth429OnSameAccount(account, http.StatusTooManyRequests, false))
 }
 
+func TestOpenAI429FastPath_DoesNotBlockOAuthWhenFallbackDisabled(t *testing.T) {
+	repo := &oauth429RateLimitRepo{}
+	settingRepo := newMockSettingRepo()
+	settingRepo.data[SettingKeyRateLimit429CooldownSettings] = `{"enabled":false,"cooldown_seconds":12}`
+	rateLimitService := NewRateLimitService(repo, nil, &config.Config{}, nil, nil)
+	rateLimitService.SetSettingService(NewSettingService(settingRepo, &config.Config{}))
+	svc := &OpenAIGatewayService{rateLimitService: rateLimitService}
+	rateLimitService.SetAccountRuntimeBlocker(svc)
+	account := &Account{ID: 425, Platform: PlatformOpenAI, Type: AccountTypeOAuth}
+	svc.openaiOAuth429RetryStartedAt.Store(account.ID, time.Now().Add(-openAIOAuth429RetryWindow-time.Second))
+
+	svc.markOpenAIOAuth429RateLimited(context.Background(), account, http.Header{}, []byte(`{"detail":"Rate limit exceeded"}`))
+
+	require.False(t, svc.isOpenAIAccountRuntimeBlocked(account), "disabled 429 fallback must not create an OAuth runtime cooldown")
+	require.Zero(t, repo.setRateLimitedCalls, "disabled 429 fallback must not persist a scheduler cooldown")
+}
+
+func TestOpenAI429FastPath_DoesNotBlockOAuthWhenQuotaWindowIsNotExhausted(t *testing.T) {
+	repo := &oauth429RateLimitRepo{}
+	settingRepo := newMockSettingRepo()
+	settingRepo.data[SettingKeyRateLimit429CooldownSettings] = `{"enabled":false,"cooldown_seconds":12}`
+	rateLimitService := NewRateLimitService(repo, nil, &config.Config{}, nil, nil)
+	rateLimitService.SetSettingService(NewSettingService(settingRepo, &config.Config{}))
+	svc := &OpenAIGatewayService{rateLimitService: rateLimitService}
+	rateLimitService.SetAccountRuntimeBlocker(svc)
+	account := &Account{ID: 426, Platform: PlatformOpenAI, Type: AccountTypeOAuth}
+	svc.openaiOAuth429RetryStartedAt.Store(account.ID, time.Now().Add(-openAIOAuth429RetryWindow-time.Second))
+	headers := http.Header{}
+	headers.Set("x-codex-primary-used-percent", "37")
+	headers.Set("x-codex-primary-reset-after-seconds", "604800")
+	headers.Set("x-codex-primary-window-minutes", "10080")
+	headers.Set("x-codex-secondary-used-percent", "20")
+	headers.Set("x-codex-secondary-reset-after-seconds", "3600")
+	headers.Set("x-codex-secondary-window-minutes", "300")
+
+	svc.handleOpenAIAccountUpstreamError(context.Background(), account, http.StatusTooManyRequests, headers, []byte(`{"detail":"Rate limit exceeded"}`))
+
+	require.False(t, svc.isOpenAIAccountRuntimeBlocked(account), "non-exhausted quota headers must use the configurable fallback")
+	require.Zero(t, repo.setRateLimitedCalls, "disabled 429 fallback must not persist a scheduler cooldown")
+}
+
 func TestOpenAI429FastPath_BlocksOAuthImmediatelyWhenSevenDayQuotaIsExhausted(t *testing.T) {
 	repo := &oauth429RateLimitRepo{}
 	rateLimits := NewRateLimitService(repo, nil, &config.Config{}, nil, nil)
@@ -307,7 +348,7 @@ func TestOpenAIHTTP429StillUsesQuotaResetHeaders(t *testing.T) {
 	account := &Account{ID: 422, Platform: PlatformOpenAI, Type: AccountTypeOAuth}
 	svc.openaiOAuth429RetryStartedAt.Store(account.ID, time.Now().Add(-openAIOAuth429RetryWindow-time.Second))
 	headers := http.Header{}
-	headers.Set("x-codex-primary-used-percent", "37")
+	headers.Set("x-codex-primary-used-percent", "100")
 	headers.Set("x-codex-primary-reset-after-seconds", "604800")
 	headers.Set("x-codex-primary-window-minutes", "10080")
 

backend/internal/service/ratelimit_service.go

@@ -1348,20 +1348,8 @@ func calculateOpenAI429ResetTime(headers http.Header) *time.Time {
 		return &resetAt
 	}
 
-	// 都未达到100%但收到429，使用较长的重置时间
-	var maxResetSecs int
-	if normalized.Reset7dSeconds != nil && *normalized.Reset7dSeconds > maxResetSecs {
-		maxResetSecs = *normalized.Reset7dSeconds
-	}
-	if normalized.Reset5hSeconds != nil && *normalized.Reset5hSeconds > maxResetSecs {
-		maxResetSecs = *normalized.Reset5hSeconds
-	}
-	if maxResetSecs > 0 {
-		resetAt := now.Add(time.Duration(maxResetSecs) * time.Second)
-		slog.Info("openai_429_using_max_reset", "max_reset_seconds", maxResetSecs, "reset_at", resetAt)
-		return &resetAt
-	}
-
+	// 未达到100%时，reset-after 只代表窗口信息，不能证明账号配额耗尽。
+	// 这类瞬时429必须回到可配置的兜底路径，避免未耗尽账号被长时间排除。
 	return nil
 }
 

backend/internal/service/ratelimit_service_openai_test.go

@@ -73,10 +73,9 @@ func TestCalculateOpenAI429ResetTime_5hExhausted(t *testing.T) {
 	}
 }
 
-func TestCalculateOpenAI429ResetTime_NeitherExhausted_UsesMax(t *testing.T) {
+func TestCalculateOpenAI429ResetTime_NeitherExhausted_ReturnsNil(t *testing.T) {
 	svc := &RateLimitService{}
 
-	// Neither limit at 100%, should use the longer reset time
 	headers := http.Header{}
 	headers.Set("x-codex-primary-used-percent", "80")
 	headers.Set("x-codex-primary-reset-after-seconds", "100000")
@@ -85,22 +84,7 @@ func TestCalculateOpenAI429ResetTime_NeitherExhausted_UsesMax(t *testing.T) {
 	headers.Set("x-codex-secondary-reset-after-seconds", "5000")
 	headers.Set("x-codex-secondary-window-minutes", "300")
 
-	before := time.Now()
-	resetAt := svc.calculateOpenAI429ResetTime(headers)
-	after := time.Now()
-
-	if resetAt == nil {
-		t.Fatal("expected non-nil resetAt")
-	}
-
-	// Should use the max (100000 seconds from 7d window)
-	expectedDuration := 100000 * time.Second
-	minExpected := before.Add(expectedDuration)
-	maxExpected := after.Add(expectedDuration)
-
-	if resetAt.Before(minExpected) || resetAt.After(maxExpected) {
-		t.Errorf("resetAt %v not in expected range [%v, %v]", resetAt, minExpected, maxExpected)
-	}
+	require.Nil(t, svc.calculateOpenAI429ResetTime(headers))
 }
 
 func TestCalculateOpenAI429ResetTime_NoCodexHeaders(t *testing.T) {