Files
easyai-ai-gateway/apps/api/internal/config/config.go
T
wangbo 3aa78d3e5c perf(worker): 按实例内存上限扩展异步容量
将策略推导出的集群目标与单 Worker 内存安全上限分离,避免失活实例的全部容量转移到单个 2 GiB Pod。

生产环境单实例上限和媒体物化并发设为 24;两实例正常时总容量由 16 提升至 48,后续可通过增加 Worker 节点继续扩展。新增 0093 迁移保存实例容量上限,并兼容滚动发布中的旧实例。

风险:更高媒体并发会增加 Worker 和节点内存压力,保留单实例 24 的硬边界并由持续监控观察 RSS、OOM、队列与节点余量。

验证:go test ./... -count=1;go vet ./...;真实 PostgreSQL 分配与故障转移测试;迁移安全检查;kubectl kustomize;gofmt -l;git diff --check。
2026-07-30 15:08:54 +08:00

363 lines
16 KiB
Go

package config
import (
"errors"
"log/slog"
"net/url"
"os"
"strconv"
"strings"
)
const (
DefaultLocalGeneratedStorageDir = "data/static/generated"
DefaultLocalUploadedStorageDir = "data/static/uploaded"
)
type Config struct {
AppEnv string
HTTPAddr string
DatabaseURL string
IdentityMode string
JWTSecret string
ServerMainBaseURL string
ServerMainInternalToken string
ServerMainInternalKey string
ServerMainInternalSecret string
IdentitySecretStore string
IdentitySecretDir string
IdentityKubernetesNamespace string
IdentityKubernetesSecretName string
IdentityKubernetesAPIServer string
IdentityKubernetesTokenFile string
IdentityKubernetesCAFile string
IdentitySecurityEventHeartbeatIntervalSeconds int
IdentitySecurityEventStaleAfterSeconds int
IdentitySecurityEventClockSkewSeconds int
PublicBaseURL string
WebBaseURL string
LocalGeneratedStorageDir string
LocalUploadedStorageDir string
LocalTempAssetTTLHours int
LocalResultTTLHours int
LocalResultMinFreeBytes int64
LocalResultMaxBytes int64
LocalResultMaxTaskBytes int64
TaskProgressCallbackEnabled bool
TaskProgressCallbackURL string
TaskProgressCallbackTimeoutMS int
TaskProgressCallbackMaxAttempts int
TaskCleanupEnabled bool
TaskRetentionDays int
TaskAnalysisRetentionDays int
TaskCleanupIntervalSeconds int
TaskCleanupBatchSize int
CORSAllowedOrigin string
GlobalHTTPProxy string
GlobalHTTPProxySource string
LogLevel slog.Level
BillingEngineMode string
ProcessRole string
DatabaseMaxConns int
MediaRequestConcurrency int
MediaMaterializationConcurrency int
AsyncQueueWorkerEnabled bool
AsyncWorkerHardLimit int
AsyncWorkerInstanceHardLimit int
AsyncWorkerRefreshIntervalSeconds int
}
func Load() Config {
globalProxy := LoadGlobalHTTPProxyStatus()
appEnv := env("APP_ENV", "development")
return Config{
AppEnv: appEnv,
HTTPAddr: env("HTTP_ADDR", ":8088"),
DatabaseURL: gatewayDatabaseURL(),
IdentityMode: env("IDENTITY_MODE", "hybrid"),
JWTSecret: env("CONFIG_JWT_SECRET", "this is a very secret secret"),
ServerMainBaseURL: strings.TrimRight(
env("SERVER_MAIN_BASE_URL", "http://localhost:3000"),
"/",
),
ServerMainInternalToken: env("SERVER_MAIN_INTERNAL_TOKEN", ""),
ServerMainInternalKey: env("SERVER_MAIN_INTERNAL_KEY", "gateway"),
ServerMainInternalSecret: env("SERVER_MAIN_INTERNAL_SECRET", env("SERVER_MAIN_INTERNAL_TOKEN", "")),
IdentitySecretStore: env("IDENTITY_SECRET_STORE", "file"),
IdentitySecretDir: env("IDENTITY_SECRET_DIR", ".local-secrets/identity"),
IdentityKubernetesNamespace: env("IDENTITY_KUBERNETES_NAMESPACE", env("POD_NAMESPACE", "")),
IdentityKubernetesSecretName: env("IDENTITY_KUBERNETES_SECRET_NAME", "easyai-gateway-identity"),
IdentityKubernetesAPIServer: env("IDENTITY_KUBERNETES_API_SERVER", "https://kubernetes.default.svc"),
IdentityKubernetesTokenFile: env("IDENTITY_KUBERNETES_TOKEN_FILE", "/var/run/secrets/kubernetes.io/serviceaccount/token"),
IdentityKubernetesCAFile: env("IDENTITY_KUBERNETES_CA_FILE", "/var/run/secrets/kubernetes.io/serviceaccount/ca.crt"),
IdentitySecurityEventHeartbeatIntervalSeconds: envInt("IDENTITY_SECURITY_EVENTS_HEARTBEAT_INTERVAL_SECONDS", 60),
IdentitySecurityEventStaleAfterSeconds: envInt("IDENTITY_SECURITY_EVENTS_STALE_AFTER_SECONDS", 180),
IdentitySecurityEventClockSkewSeconds: envInt("IDENTITY_SECURITY_EVENTS_CLOCK_SKEW_SECONDS", 60),
PublicBaseURL: strings.TrimRight(env("AI_GATEWAY_PUBLIC_BASE_URL", env("PUBLIC_BASE_URL", "")), "/"),
WebBaseURL: strings.TrimRight(env("AI_GATEWAY_WEB_BASE_URL", env("GATEWAY_WEB_BASE_URL", env("PUBLIC_WEB_BASE_URL", ""))), "/"),
LocalGeneratedStorageDir: env("AI_GATEWAY_GENERATED_STORAGE_DIR", env("LOCAL_GENERATED_STORAGE_DIR", env("AI_GATEWAY_STATIC_STORAGE_DIR", DefaultLocalGeneratedStorageDir))),
LocalUploadedStorageDir: env("AI_GATEWAY_UPLOADED_STORAGE_DIR", env("LOCAL_UPLOADED_STORAGE_DIR", DefaultLocalUploadedStorageDir)),
LocalTempAssetTTLHours: envInt("AI_GATEWAY_LOCAL_TEMP_ASSET_TTL_HOURS", 24),
LocalResultTTLHours: envIntValidated("AI_GATEWAY_LOCAL_RESULT_TTL_HOURS", 24),
LocalResultMinFreeBytes: envInt64Validated("AI_GATEWAY_LOCAL_RESULT_MIN_FREE_BYTES", 10*1024*1024*1024),
LocalResultMaxBytes: envInt64Validated("AI_GATEWAY_LOCAL_RESULT_MAX_BYTES", 256*1024*1024),
LocalResultMaxTaskBytes: envInt64Validated("AI_GATEWAY_LOCAL_RESULT_MAX_TASK_BYTES", 512*1024*1024),
TaskProgressCallbackEnabled: env("TASK_PROGRESS_CALLBACK_ENABLED", "true") == "true",
TaskProgressCallbackURL: env("TASK_PROGRESS_CALLBACK_URL",
strings.TrimRight(env("SERVER_MAIN_BASE_URL", "http://localhost:3000"), "/")+"/internal/platform/task-progress-callbacks",
),
TaskProgressCallbackTimeoutMS: envIntValidated("TASK_PROGRESS_CALLBACK_TIMEOUT_MS", 5000),
TaskProgressCallbackMaxAttempts: envIntValidated("TASK_PROGRESS_CALLBACK_MAX_ATTEMPTS", 10),
TaskCleanupEnabled: env("AI_GATEWAY_TASK_CLEANUP_ENABLED", "false") == "true",
TaskRetentionDays: envIntValidated("AI_GATEWAY_TASK_RETENTION_DAYS", 30),
TaskAnalysisRetentionDays: envIntValidated("AI_GATEWAY_TASK_ANALYSIS_RETENTION_DAYS", 7),
TaskCleanupIntervalSeconds: envIntValidated("AI_GATEWAY_TASK_CLEANUP_INTERVAL_SECONDS", 300),
TaskCleanupBatchSize: envIntValidated("AI_GATEWAY_TASK_CLEANUP_BATCH_SIZE", 1000),
CORSAllowedOrigin: env("CORS_ALLOWED_ORIGIN", "http://localhost:5178,http://127.0.0.1:5178"),
GlobalHTTPProxy: globalProxy.HTTPProxy,
GlobalHTTPProxySource: globalProxy.Source,
LogLevel: logLevel(env("LOG_LEVEL", "info")),
BillingEngineMode: strings.ToLower(env("BILLING_ENGINE_MODE", "observe")),
ProcessRole: strings.ToLower(strings.TrimSpace(env("AI_GATEWAY_PROCESS_ROLE", "all"))),
DatabaseMaxConns: envInt("AI_GATEWAY_DATABASE_MAX_CONNS", 0),
MediaRequestConcurrency: envIntValidated("AI_GATEWAY_MEDIA_REQUEST_CONCURRENCY", 16),
MediaMaterializationConcurrency: envIntValidated("AI_GATEWAY_MEDIA_MATERIALIZATION_CONCURRENCY", 8),
AsyncQueueWorkerEnabled: env("AI_GATEWAY_ASYNC_QUEUE_WORKER_ENABLED", "true") == "true",
AsyncWorkerHardLimit: envIntValidated("AI_GATEWAY_ASYNC_WORKER_HARD_LIMIT", 2048),
AsyncWorkerInstanceHardLimit: envIntValidated("AI_GATEWAY_ASYNC_WORKER_INSTANCE_HARD_LIMIT", 32),
AsyncWorkerRefreshIntervalSeconds: envIntValidated("AI_GATEWAY_ASYNC_WORKER_REFRESH_INTERVAL_SECONDS", 5),
}
}
func (c Config) Validate() error {
switch strings.ToLower(strings.TrimSpace(c.ProcessRole)) {
case "", "all", "api", "worker":
default:
return errors.New("AI_GATEWAY_PROCESS_ROLE must be all, api, or worker")
}
if c.DatabaseMaxConns < 0 || c.DatabaseMaxConns > 1000 {
return errors.New("AI_GATEWAY_DATABASE_MAX_CONNS must be between 1 and 1000 when configured")
}
if c.MediaMaterializationConcurrency != 0 && (c.MediaMaterializationConcurrency < 1 || c.MediaMaterializationConcurrency > 256) {
return errors.New("AI_GATEWAY_MEDIA_MATERIALIZATION_CONCURRENCY must be between 1 and 256")
}
if c.MediaRequestConcurrency != 0 && (c.MediaRequestConcurrency < 1 || c.MediaRequestConcurrency > 1024) {
return errors.New("AI_GATEWAY_MEDIA_REQUEST_CONCURRENCY must be between 1 and 1024")
}
switch strings.ToLower(strings.TrimSpace(c.BillingEngineMode)) {
case "", "observe", "enforce", "hold":
default:
return errors.New("BILLING_ENGINE_MODE must be observe, enforce, or hold")
}
if c.AsyncWorkerHardLimit < 1 || c.AsyncWorkerHardLimit > 10000 {
return errors.New("AI_GATEWAY_ASYNC_WORKER_HARD_LIMIT must be between 1 and 10000")
}
if c.AsyncWorkerInstanceHardLimit < 1 || c.AsyncWorkerInstanceHardLimit > 10000 {
return errors.New("AI_GATEWAY_ASYNC_WORKER_INSTANCE_HARD_LIMIT must be between 1 and 10000")
}
if c.AsyncWorkerRefreshIntervalSeconds < 1 {
return errors.New("AI_GATEWAY_ASYNC_WORKER_REFRESH_INTERVAL_SECONDS must be positive")
}
if c.TaskProgressCallbackTimeoutMS != 0 && (c.TaskProgressCallbackTimeoutMS < 100 || c.TaskProgressCallbackTimeoutMS > 60000) {
return errors.New("TASK_PROGRESS_CALLBACK_TIMEOUT_MS must be between 100 and 60000")
}
if c.TaskProgressCallbackMaxAttempts != 0 && (c.TaskProgressCallbackMaxAttempts < 1 || c.TaskProgressCallbackMaxAttempts > 100) {
return errors.New("TASK_PROGRESS_CALLBACK_MAX_ATTEMPTS must be between 1 and 100")
}
if c.TaskRetentionDays != 0 && (c.TaskRetentionDays < 1 || c.TaskRetentionDays > 3650) {
return errors.New("AI_GATEWAY_TASK_RETENTION_DAYS must be between 1 and 3650")
}
if c.TaskAnalysisRetentionDays != 0 && (c.TaskAnalysisRetentionDays < 1 || (c.TaskRetentionDays != 0 && c.TaskAnalysisRetentionDays > c.TaskRetentionDays)) {
return errors.New("AI_GATEWAY_TASK_ANALYSIS_RETENTION_DAYS must be between 1 and AI_GATEWAY_TASK_RETENTION_DAYS")
}
if c.TaskCleanupIntervalSeconds != 0 && c.TaskCleanupIntervalSeconds < 60 {
return errors.New("AI_GATEWAY_TASK_CLEANUP_INTERVAL_SECONDS must be at least 60")
}
if c.TaskCleanupBatchSize != 0 && (c.TaskCleanupBatchSize < 100 || c.TaskCleanupBatchSize > 5000) {
return errors.New("AI_GATEWAY_TASK_CLEANUP_BATCH_SIZE must be between 100 and 5000")
}
if c.LocalResultTTLHours != 0 && (c.LocalResultTTLHours < 1 || c.LocalResultTTLHours > 24*30) {
return errors.New("AI_GATEWAY_LOCAL_RESULT_TTL_HOURS must be between 1 and 720")
}
if c.LocalResultMinFreeBytes < 0 {
return errors.New("AI_GATEWAY_LOCAL_RESULT_MIN_FREE_BYTES must not be negative")
}
if c.LocalResultMaxBytes != 0 && c.LocalResultMaxBytes < 1024 {
return errors.New("AI_GATEWAY_LOCAL_RESULT_MAX_BYTES must be at least 1024")
}
if c.LocalResultMaxTaskBytes != 0 && c.LocalResultMaxTaskBytes < c.LocalResultMaxBytes {
return errors.New("AI_GATEWAY_LOCAL_RESULT_MAX_TASK_BYTES must be at least AI_GATEWAY_LOCAL_RESULT_MAX_BYTES")
}
switch strings.ToLower(strings.TrimSpace(c.IdentitySecretStore)) {
case "":
case "file":
if strings.TrimSpace(c.IdentitySecretDir) == "" {
return errors.New("IDENTITY_SECRET_DIR is required for the file SecretStore")
}
case "kubernetes":
if strings.TrimSpace(c.IdentityKubernetesNamespace) == "" || strings.TrimSpace(c.IdentityKubernetesSecretName) == "" {
return errors.New("Kubernetes identity SecretStore requires namespace and Secret name")
}
default:
return errors.New("IDENTITY_SECRET_STORE must be file or kubernetes")
}
if c.IdentitySecurityEventHeartbeatIntervalSeconds != 0 || c.IdentitySecurityEventStaleAfterSeconds != 0 || c.IdentitySecurityEventClockSkewSeconds != 0 {
if c.IdentitySecurityEventHeartbeatIntervalSeconds <= 0 ||
c.IdentitySecurityEventStaleAfterSeconds < 2*c.IdentitySecurityEventHeartbeatIntervalSeconds ||
c.IdentitySecurityEventClockSkewSeconds < 0 || c.IdentitySecurityEventClockSkewSeconds > 300 {
return errors.New("identity security event heartbeat, stale threshold, or clock skew is invalid")
}
}
return nil
}
func (c Config) EffectiveProcessRole() string {
role := strings.ToLower(strings.TrimSpace(c.ProcessRole))
if role == "" {
return "all"
}
return role
}
func (c Config) RunsPublicHTTP() bool {
return c.EffectiveProcessRole() != "worker"
}
func (c Config) RunsAsyncExecutionWorker() bool {
switch c.EffectiveProcessRole() {
case "api":
return false
case "worker":
return true
default:
return c.AsyncQueueWorkerEnabled
}
}
func (c Config) RunsBackgroundWorkers() bool {
return c.EffectiveProcessRole() != "api"
}
type GlobalHTTPProxyStatus struct {
HTTPProxy string
Source string
}
func LoadGlobalHTTPProxyStatus() GlobalHTTPProxyStatus {
for _, key := range []string{
"AI_GATEWAY_GLOBAL_HTTP_PROXY",
"GLOBAL_HTTP_PROXY",
"HTTPS_PROXY",
"https_proxy",
"HTTP_PROXY",
"http_proxy",
"ALL_PROXY",
"all_proxy",
} {
if value := envValue(key); value != "" {
return GlobalHTTPProxyStatus{HTTPProxy: value, Source: key}
}
}
return GlobalHTTPProxyStatus{}
}
func gatewayDatabaseURL() string {
if value := envValue("AI_GATEWAY_DATABASE_URL"); value != "" {
return normalizePostgresURL(value)
}
if value := envValue("DATABASE_URL"); value != "" {
return normalizePostgresURL(value)
}
if memoryURL := envValue("MEMORY_DATABASE_URL"); memoryURL != "" {
return normalizePostgresURL(withDatabase(memoryURL, env("AI_GATEWAY_DATABASE_NAME", "easyai_ai_gateway")))
}
return normalizePostgresURL("postgresql://easyai:easyai2025@localhost:5432/easyai_ai_gateway?sslmode=disable")
}
func normalizePostgresURL(raw string) string {
parsed, err := url.Parse(raw)
if err != nil {
return raw
}
values := parsed.Query()
schema := values.Get("schema")
if schema == "" {
return raw
}
values.Del("schema")
if values.Get("search_path") == "" {
values.Set("search_path", schema)
}
parsed.RawQuery = values.Encode()
return parsed.String()
}
func withDatabase(raw string, databaseName string) string {
parsed, err := url.Parse(raw)
if err != nil || databaseName == "" {
return raw
}
parsed.Path = "/" + databaseName
return parsed.String()
}
func envValue(key string) string {
return strings.TrimSpace(os.Getenv(key))
}
func env(key string, fallback string) string {
if value := envValue(key); value != "" {
return value
}
return fallback
}
func envInt(key string, fallback int) int {
value := envValue(key)
if value == "" {
return fallback
}
parsed, err := strconv.Atoi(value)
if err != nil {
return fallback
}
return parsed
}
func envIntValidated(key string, fallback int) int {
value := envValue(key)
if value == "" {
return fallback
}
parsed, err := strconv.Atoi(value)
if err != nil {
return 0
}
return parsed
}
func envInt64Validated(key string, fallback int64) int64 {
value := envValue(key)
if value == "" {
return fallback
}
parsed, err := strconv.ParseInt(value, 10, 64)
if err != nil {
return 0
}
return parsed
}
func logLevel(value string) slog.Level {
switch strings.ToLower(value) {
case "debug":
return slog.LevelDebug
case "warn", "warning":
return slog.LevelWarn
case "error":
return slog.LevelError
default:
return slog.LevelInfo
}
}