fork-it/gpt-load
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

feat: 优化超时配置

Browse Source
This commit is contained in:
tbphp
2025-06-13 12:36:13 +08:00
parent 3232d34534
commit d82d46376f
8 changed files with 137 additions and 80 deletions
Download Patch File Download Diff File Expand all files Collapse all files

27
internal/config/manager.go
View File

@@ -30,8 +30,8 @@ type Constants struct {
var DefaultConstants = Constants{
MinPort: 1,
MaxPort: 65535,
MinTimeout: 1000,
DefaultTimeout: 30000,
MinTimeout: 1,
DefaultTimeout: 30,
DefaultMaxSockets: 50,
DefaultMaxFreeSockets: 10,
}
@@ -62,8 +62,12 @@ func NewManager() (types.ConfigManager, error) {
config := &Config{
Server: types.ServerConfig{
Port: parseInteger(os.Getenv("PORT"), 3000),
Host: getEnvOrDefault("HOST", "0.0.0.0"),
Port: parseInteger(os.Getenv("PORT"), 3000),
Host: getEnvOrDefault("HOST", "0.0.0.0"),
ReadTimeout: parseInteger(os.Getenv("SERVER_READ_TIMEOUT"), 120),
WriteTimeout: parseInteger(os.Getenv("SERVER_WRITE_TIMEOUT"), 1800),
IdleTimeout: parseInteger(os.Getenv("SERVER_IDLE_TIMEOUT"), 120),
GracefulShutdownTimeout: parseInteger(os.Getenv("SERVER_GRACEFUL_SHUTDOWN_TIMEOUT"), 60),
},
Keys: types.KeysConfig{
FilePath: getEnvOrDefault("KEYS_FILE", "keys.txt"),
@@ -72,8 +76,10 @@ func NewManager() (types.ConfigManager, error) {
MaxRetries: parseInteger(os.Getenv("MAX_RETRIES"), 3),
},
OpenAI: types.OpenAIConfig{
BaseURLs: parseArray(os.Getenv("OPENAI_BASE_URL"), []string{"https://api.openai.com"}),
Timeout: parseInteger(os.Getenv("REQUEST_TIMEOUT"), DefaultConstants.DefaultTimeout),
BaseURLs: parseArray(os.Getenv("OPENAI_BASE_URL"), []string{"https://api.openai.com"}),
RequestTimeout: parseInteger(os.Getenv("REQUEST_TIMEOUT"), DefaultConstants.DefaultTimeout),
ResponseTimeout: parseInteger(os.Getenv("RESPONSE_TIMEOUT"), 30),
IdleConnTimeout: parseInteger(os.Getenv("IDLE_CONN_TIMEOUT"), 120),
},
Auth: types.AuthConfig{
Key: os.Getenv("AUTH_KEY"),
@@ -88,7 +94,6 @@ func NewManager() (types.ConfigManager, error) {
},
Performance: types.PerformanceConfig{
MaxConcurrentRequests: parseInteger(os.Getenv("MAX_CONCURRENT_REQUESTS"), 100),
RequestTimeout: parseInteger(os.Getenv("REQUEST_TIMEOUT"), DefaultConstants.DefaultTimeout),
EnableGzip: parseBoolean(os.Getenv("ENABLE_GZIP"), true),
},
Log: types.LogConfig{
@@ -173,8 +178,8 @@ func (m *Manager) Validate() error {
}
// Validate timeout
if m.config.OpenAI.Timeout < DefaultConstants.MinTimeout {
validationErrors = append(validationErrors, fmt.Sprintf("request timeout cannot be less than %dms", DefaultConstants.MinTimeout))
if m.config.OpenAI.RequestTimeout < DefaultConstants.MinTimeout {
validationErrors = append(validationErrors, fmt.Sprintf("request timeout cannot be less than %ds", DefaultConstants.MinTimeout))
}
// Validate upstream URL format
@@ -212,7 +217,9 @@ func (m *Manager) DisplayConfig() {
logrus.Infof(" Blacklist threshold: %d errors", m.config.Keys.BlacklistThreshold)
logrus.Infof(" Max retries: %d", m.config.Keys.MaxRetries)
logrus.Infof(" Upstream URLs: %s", strings.Join(m.config.OpenAI.BaseURLs, ", "))
logrus.Infof(" Request timeout: %dms", m.config.OpenAI.Timeout)
logrus.Infof(" Request timeout: %ds", m.config.OpenAI.RequestTimeout)
logrus.Infof(" Response timeout: %ds", m.config.OpenAI.ResponseTimeout)
logrus.Infof(" Idle connection timeout: %ds", m.config.OpenAI.IdleConnTimeout)
authStatus := "disabled"
if m.config.Auth.Enabled {

16
internal/handler/handler.go
View File

@@ -176,8 +176,10 @@ func (h *Handler) GetConfig(c *gin.Context) {
"max_retries": keysConfig.MaxRetries,
},
"openai": gin.H{
"base_url": openaiConfig.BaseURL,
"timeout": openaiConfig.Timeout,
"base_url": openaiConfig.BaseURL,
"request_timeout": openaiConfig.RequestTimeout,
"response_timeout": openaiConfig.ResponseTimeout,
"idle_conn_timeout": openaiConfig.IdleConnTimeout,
},
"auth": gin.H{
"enabled": authConfig.Enabled,
@@ -192,9 +194,17 @@ func (h *Handler) GetConfig(c *gin.Context) {
},
"performance": gin.H{
"max_concurrent_requests": perfConfig.MaxConcurrentRequests,
"request_timeout": perfConfig.RequestTimeout,
"enable_gzip": perfConfig.EnableGzip,
},
"timeout_config": gin.H{
"request_timeout_s": openaiConfig.RequestTimeout,
"response_timeout_s": openaiConfig.ResponseTimeout,
"idle_conn_timeout_s": openaiConfig.IdleConnTimeout,
"server_read_timeout_s": serverConfig.ReadTimeout,
"server_write_timeout_s": serverConfig.WriteTimeout,
"server_idle_timeout_s": serverConfig.IdleTimeout,
"graceful_shutdown_timeout_s": serverConfig.GracefulShutdownTimeout,
},
"log": gin.H{
"level": logConfig.Level,
"format": logConfig.Format,

15
internal/proxy/server.go
View File

@@ -53,14 +53,13 @@ func NewProxyServer(keyManager types.KeyManager, configManager types.ConfigManag
openaiConfig := configManager.GetOpenAIConfig()
perfConfig := configManager.GetPerformanceConfig()
// Create high-performance HTTP client
transport := &http.Transport{
MaxIdleConns: 100,
MaxIdleConnsPerHost: 20,
MaxConnsPerHost: 0,
IdleConnTimeout: 120 * time.Second,
TLSHandshakeTimeout: 30 * time.Second,
IdleConnTimeout: time.Duration(openaiConfig.IdleConnTimeout) * time.Second,
TLSHandshakeTimeout: time.Duration(openaiConfig.ResponseTimeout) * time.Second,
ExpectContinueTimeout: 1 * time.Second,
DisableCompression: !perfConfig.EnableGzip,
ForceAttemptHTTP2: true,
@@ -73,19 +72,19 @@ func NewProxyServer(keyManager types.KeyManager, configManager types.ConfigManag
MaxIdleConns: 200,
MaxIdleConnsPerHost: 40,
MaxConnsPerHost: 0,
IdleConnTimeout: 300 * time.Second, // Keep streaming connections longer
TLSHandshakeTimeout: 30 * time.Second,
IdleConnTimeout: time.Duration(openaiConfig.IdleConnTimeout) * time.Second,
TLSHandshakeTimeout: time.Duration(openaiConfig.ResponseTimeout) * time.Second,
ExpectContinueTimeout: 1 * time.Second,
DisableCompression: true, // Always disable compression for streaming
ForceAttemptHTTP2: true,
WriteBufferSize: 64 * 1024,
ReadBufferSize: 64 * 1024,
ResponseHeaderTimeout: 30 * time.Second,
ResponseHeaderTimeout: time.Duration(openaiConfig.ResponseTimeout) * time.Second,
}
httpClient := &http.Client{
Transport: transport,
Timeout: time.Duration(openaiConfig.Timeout) * time.Millisecond,
Timeout: time.Duration(openaiConfig.RequestTimeout) * time.Second,
}
// Streaming client without overall timeout
@@ -229,7 +228,7 @@ func (ps *ProxyServer) executeRequestWithRetry(c *gin.Context, startTime time.Ti
ctx, cancel = context.WithCancel(c.Request.Context())
} else {
// Non-streaming requests use configured timeout from the already fetched config
timeout := time.Duration(openaiConfig.Timeout) * time.Millisecond
timeout := time.Duration(openaiConfig.RequestTimeout) * time.Second
ctx, cancel = context.WithTimeout(c.Request.Context(), timeout)
}
defer cancel()