Files
easyai-ai-gateway/apps/api/internal/store/leadership.go
T
wangbo 7786692d32 feat(routing): 引入多执行池智能调度
将 Worker 发现、路由画像、容量与执行传输抽象为平台无关接口,新增 Kubernetes 和静态容量适配器,并以 shadow 模式接入生产配置。

实现网络与容量评分、路由防抖、池队列、同步 Worker 租约、一次性执行令牌,以及提交状态不明时禁止重复分配的安全语义。

新增 0105 兼容迁移、管理接口、指标、OpenAPI 和回归测试。已执行全量 Go 测试、go vet、OpenAPI、迁移安全、Compose 与 Kustomize 验证。
2026-08-05 22:25:37 +08:00

103 lines
3.0 KiB
Go

package store
import (
"context"
"errors"
"sync"
"time"
"github.com/jackc/pgx/v5/pgxpool"
)
const leadershipReleaseTimeout = 5 * time.Second
const (
identityCoordinatorLeadershipKey = "easyai-gateway-identity-coordinator"
securityEventHeartbeatLeadershipKey = "easyai-gateway-security-event-heartbeat"
capacityControllerLeadershipKey = "easyai-gateway-capacity-controller"
)
// Leadership holds a PostgreSQL session advisory lock. Callers must keep the
// lease alive and release it when leadership is no longer needed.
type Leadership interface {
KeepAlive(context.Context) error
Release()
}
type postgresLeadership struct {
conn *pgxpool.Conn
key string
once sync.Once
}
func (s *Store) TryAcquireIdentityCoordinatorLeadership(ctx context.Context) (Leadership, bool, error) {
return s.tryAcquireLeadership(ctx, identityCoordinatorLeadershipKey)
}
func (s *Store) TryAcquireSecurityEventHeartbeatLeadership(ctx context.Context) (Leadership, bool, error) {
return s.tryAcquireLeadership(ctx, securityEventHeartbeatLeadershipKey)
}
func (s *Store) TryAcquireCapacityControllerLeadership(ctx context.Context) (Leadership, bool, error) {
return s.tryAcquireLeadership(ctx, capacityControllerLeadershipKey)
}
func (s *Store) TryAcquireRouteProbeLeadership(ctx context.Context, poolID, routeProfileKey string) (Leadership, bool, error) {
return s.tryAcquireLeadership(ctx, "easyai-gateway-route-probe:"+poolID+":"+routeProfileKey)
}
func (s *Store) tryAcquireLeadership(ctx context.Context, key string) (Leadership, bool, error) {
conn, err := s.pool.Acquire(ctx)
if err != nil {
return nil, false, err
}
var acquired bool
if err := conn.QueryRow(ctx,
`SELECT pg_try_advisory_lock(hashtextextended($1, 0))`,
key,
).Scan(&acquired); err != nil {
conn.Release()
return nil, false, err
}
if !acquired {
conn.Release()
return nil, false, nil
}
return &postgresLeadership{conn: conn, key: key}, true, nil
}
func (leadership *postgresLeadership) KeepAlive(ctx context.Context) error {
if leadership == nil || leadership.conn == nil {
return errors.New("leadership lease is closed")
}
var alive int
return leadership.conn.QueryRow(ctx, `SELECT 1`).Scan(&alive)
}
func (leadership *postgresLeadership) Release() {
if leadership == nil {
return
}
leadership.once.Do(func() {
unlockCtx, unlockCancel := context.WithTimeout(context.Background(), leadershipReleaseTimeout)
var unlocked bool
err := leadership.conn.QueryRow(unlockCtx,
`SELECT pg_advisory_unlock(hashtextextended($1, 0))`,
leadership.key,
).Scan(&unlocked)
unlockCancel()
if err != nil || !unlocked {
// A session-scoped advisory lock must never return to the pool when
// unlock is uncertain. Closing the connection lets PostgreSQL
// release the lock and forces the pool to replace the session.
conn := leadership.conn.Hijack()
closeCtx, closeCancel := context.WithTimeout(context.Background(), leadershipReleaseTimeout)
_ = conn.Close(closeCtx)
closeCancel()
} else {
leadership.conn.Release()
}
leadership.conn = nil
})
}