feat(gateway): 实现负载感知的账号调度优化
- 新增调度配置：粘性会话排队、兜底排队、负载计算、槽位清理 - 实现账号级等待队列和批量负载查询（Redis Lua 脚本） - 三层选择策略：粘性会话优先 → 负载感知选择 → 兜底排队 - 后台定期清理过期槽位，防止资源泄漏 - 集成到所有网关处理器（Claude/Gemini/OpenAI）
This commit is contained in:
IanShaw027
@@ -141,6 +141,10 @@ func (h *GatewayHandler) Messages(c *gin.Context) {
|
||||
} else if apiKey.Group != nil {
|
||||
platform = apiKey.Group.Platform
|
||||
}
|
||||
sessionKey := sessionHash
|
||||
if platform == service.PlatformGemini && sessionHash != "" {
|
||||
sessionKey = "gemini:" + sessionHash
|
||||
}
|
||||
|
||||
if platform == service.PlatformGemini {
|
||||
const maxAccountSwitches = 3
|
||||
@@ -149,7 +153,7 @@ func (h *GatewayHandler) Messages(c *gin.Context) {
|
||||
lastFailoverStatus := 0
|
||||
|
||||
for {
|
||||
account, err := h.geminiCompatService.SelectAccountForModelWithExclusions(c.Request.Context(), apiKey.GroupID, sessionHash, reqModel, failedAccountIDs)
|
||||
selection, err := h.gatewayService.SelectAccountWithLoadAwareness(c.Request.Context(), apiKey.GroupID, sessionKey, reqModel, failedAccountIDs)
|
||||
if err != nil {
|
||||
if len(failedAccountIDs) == 0 {
|
||||
h.handleStreamingAwareError(c, http.StatusServiceUnavailable, "api_error", "No available accounts: "+err.Error(), streamStarted)
|
||||
@@ -158,9 +162,13 @@ func (h *GatewayHandler) Messages(c *gin.Context) {
|
||||
h.handleFailoverExhausted(c, lastFailoverStatus, streamStarted)
|
||||
return
|
||||
}
|
||||
account := selection.Account
|
||||
|
||||
// 检查预热请求拦截（在账号选择后、转发前检查）
|
||||
if account.IsInterceptWarmupEnabled() && isWarmupRequest(body) {
|
||||
if selection.Acquired && selection.ReleaseFunc != nil {
|
||||
selection.ReleaseFunc()
|
||||
}
|
||||
if reqStream {
|
||||
sendMockWarmupStream(c, reqModel)
|
||||
} else {
|
||||
@@ -170,11 +178,44 @@ func (h *GatewayHandler) Messages(c *gin.Context) {
|
||||
}
|
||||
|
||||
// 3. 获取账号并发槽位
|
||||
accountReleaseFunc, err := h.concurrencyHelper.AcquireAccountSlotWithWait(c, account.ID, account.Concurrency, reqStream, &streamStarted)
|
||||
if err != nil {
|
||||
log.Printf("Account concurrency acquire failed: %v", err)
|
||||
h.handleConcurrencyError(c, err, "account", streamStarted)
|
||||
return
|
||||
accountReleaseFunc := selection.ReleaseFunc
|
||||
var accountWaitRelease func()
|
||||
if !selection.Acquired {
|
||||
if selection.WaitPlan == nil {
|
||||
h.handleStreamingAwareError(c, http.StatusServiceUnavailable, "api_error", "No available accounts", streamStarted)
|
||||
return
|
||||
}
|
||||
canWait, err := h.concurrencyHelper.IncrementAccountWaitCount(c.Request.Context(), account.ID, selection.WaitPlan.MaxWaiting)
|
||||
if err != nil {
|
||||
log.Printf("Increment account wait count failed: %v", err)
|
||||
} else if !canWait {
|
||||
log.Printf("Account wait queue full: account=%d", account.ID)
|
||||
h.handleStreamingAwareError(c, http.StatusTooManyRequests, "rate_limit_error", "Too many pending requests, please retry later", streamStarted)
|
||||
return
|
||||
}
|
||||
accountWaitRelease = func() {
|
||||
h.concurrencyHelper.DecrementAccountWaitCount(c.Request.Context(), account.ID)
|
||||
}
|
||||
|
||||
accountReleaseFunc, err = h.concurrencyHelper.AcquireAccountSlotWithWaitTimeout(
|
||||
c,
|
||||
account.ID,
|
||||
selection.WaitPlan.MaxConcurrency,
|
||||
selection.WaitPlan.Timeout,
|
||||
reqStream,
|
||||
&streamStarted,
|
||||
)
|
||||
if err != nil {
|
||||
if accountWaitRelease != nil {
|
||||
accountWaitRelease()
|
||||
}
|
||||
log.Printf("Account concurrency acquire failed: %v", err)
|
||||
h.handleConcurrencyError(c, err, "account", streamStarted)
|
||||
return
|
||||
}
|
||||
if err := h.gatewayService.BindStickySession(c.Request.Context(), sessionKey, account.ID); err != nil {
|
||||
log.Printf("Bind sticky session failed: %v", err)
|
||||
}
|
||||
}
|
||||
|
||||
// 转发请求 - 根据账号平台分流
|
||||
@@ -187,6 +228,9 @@ func (h *GatewayHandler) Messages(c *gin.Context) {
|
||||
if accountReleaseFunc != nil {
|
||||
accountReleaseFunc()
|
||||
}
|
||||
if accountWaitRelease != nil {
|
||||
accountWaitRelease()
|
||||
}
|
||||
if err != nil {
|
||||
var failoverErr *service.UpstreamFailoverError
|
||||
if errors.As(err, &failoverErr) {
|
||||
@@ -231,7 +275,7 @@ func (h *GatewayHandler) Messages(c *gin.Context) {
|
||||
|
||||
for {
|
||||
// 选择支持该模型的账号
|
||||
account, err := h.gatewayService.SelectAccountForModelWithExclusions(c.Request.Context(), apiKey.GroupID, sessionHash, reqModel, failedAccountIDs)
|
||||
selection, err := h.gatewayService.SelectAccountWithLoadAwareness(c.Request.Context(), apiKey.GroupID, sessionKey, reqModel, failedAccountIDs)
|
||||
if err != nil {
|
||||
if len(failedAccountIDs) == 0 {
|
||||
h.handleStreamingAwareError(c, http.StatusServiceUnavailable, "api_error", "No available accounts: "+err.Error(), streamStarted)
|
||||
@@ -240,9 +284,13 @@ func (h *GatewayHandler) Messages(c *gin.Context) {
|
||||
h.handleFailoverExhausted(c, lastFailoverStatus, streamStarted)
|
||||
return
|
||||
}
|
||||
account := selection.Account
|
||||
|
||||
// 检查预热请求拦截（在账号选择后、转发前检查）
|
||||
if account.IsInterceptWarmupEnabled() && isWarmupRequest(body) {
|
||||
if selection.Acquired && selection.ReleaseFunc != nil {
|
||||
selection.ReleaseFunc()
|
||||
}
|
||||
if reqStream {
|
||||
sendMockWarmupStream(c, reqModel)
|
||||
} else {
|
||||
@@ -252,11 +300,44 @@ func (h *GatewayHandler) Messages(c *gin.Context) {
|
||||
}
|
||||
|
||||
// 3. 获取账号并发槽位
|
||||
accountReleaseFunc, err := h.concurrencyHelper.AcquireAccountSlotWithWait(c, account.ID, account.Concurrency, reqStream, &streamStarted)
|
||||
if err != nil {
|
||||
log.Printf("Account concurrency acquire failed: %v", err)
|
||||
h.handleConcurrencyError(c, err, "account", streamStarted)
|
||||
return
|
||||
accountReleaseFunc := selection.ReleaseFunc
|
||||
var accountWaitRelease func()
|
||||
if !selection.Acquired {
|
||||
if selection.WaitPlan == nil {
|
||||
h.handleStreamingAwareError(c, http.StatusServiceUnavailable, "api_error", "No available accounts", streamStarted)
|
||||
return
|
||||
}
|
||||
canWait, err := h.concurrencyHelper.IncrementAccountWaitCount(c.Request.Context(), account.ID, selection.WaitPlan.MaxWaiting)
|
||||
if err != nil {
|
||||
log.Printf("Increment account wait count failed: %v", err)
|
||||
} else if !canWait {
|
||||
log.Printf("Account wait queue full: account=%d", account.ID)
|
||||
h.handleStreamingAwareError(c, http.StatusTooManyRequests, "rate_limit_error", "Too many pending requests, please retry later", streamStarted)
|
||||
return
|
||||
}
|
||||
accountWaitRelease = func() {
|
||||
h.concurrencyHelper.DecrementAccountWaitCount(c.Request.Context(), account.ID)
|
||||
}
|
||||
|
||||
accountReleaseFunc, err = h.concurrencyHelper.AcquireAccountSlotWithWaitTimeout(
|
||||
c,
|
||||
account.ID,
|
||||
selection.WaitPlan.MaxConcurrency,
|
||||
selection.WaitPlan.Timeout,
|
||||
reqStream,
|
||||
&streamStarted,
|
||||
)
|
||||
if err != nil {
|
||||
if accountWaitRelease != nil {
|
||||
accountWaitRelease()
|
||||
}
|
||||
log.Printf("Account concurrency acquire failed: %v", err)
|
||||
h.handleConcurrencyError(c, err, "account", streamStarted)
|
||||
return
|
||||
}
|
||||
if err := h.gatewayService.BindStickySession(c.Request.Context(), sessionKey, account.ID); err != nil {
|
||||
log.Printf("Bind sticky session failed: %v", err)
|
||||
}
|
||||
}
|
||||
|
||||
// 转发请求 - 根据账号平台分流
|
||||
@@ -269,6 +350,9 @@ func (h *GatewayHandler) Messages(c *gin.Context) {
|
||||
if accountReleaseFunc != nil {
|
||||
accountReleaseFunc()
|
||||
}
|
||||
if accountWaitRelease != nil {
|
||||
accountWaitRelease()
|
||||
}
|
||||
if err != nil {
|
||||
var failoverErr *service.UpstreamFailoverError
|
||||
if errors.As(err, &failoverErr) {
|
||||
|
||||
Reference in New Issue
Block a user