Files
pangolin/server/internal/scheduler/wiring.go
T
wangjia 5f3f4189e5 feat(scheduler): lifecycle 真实接线适配器(2/2) + 装配 BuildRealConfig
把 scheduler 的 stub lifecycle 换成 SQL 后端真实实现,并接入 server:
- SQLLifecycle:MySQL 后端,同时满足 detect.LifecycleService 与
  orchestrate.LifecycleService(两个 thin adapter 包装共享核心)。
  - TransitionStatus:乐观锁 UPDATE nodes SET status WHERE id AND status=from
    → rows affected(1=成功/0=冲突 no-op)+ 写 node_events(按状态映射 enum,
    eventForStatus 纯函数已单测)+ bump directory_version,全在一个事务内。
  - ListNodes/GetNode/SetWeight/BumpVersion/GetLoad(读 LoadCache)/WriteAuditLog。
  - 已知限制:无历史 load 表 → GetLoadHistory 返回当前点(掉量规则安全降级);
    destroy 的 cert 撤销留 TODO(provision 已拆 VM,撤销为纵深防御,待机群再接)。
- BuildRealConfig:用真实 lifecycle + provisionAdapter 装配三循环。
- main.go:SCHED_ENABLED 且有 DB 时走 BuildRealConfig(provision 无厂商凭证则
  CreateNode 优雅失败、替换保持 pending),否则回退 stub。
- 全量 server 23 包测试通过;e2e(判封→drain→换机→置备)待机群+厂商凭证验证。

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2026-06-17 09:20:40 +08:00

163 lines
6.3 KiB
Go

package scheduler
// wiring.go — construction helpers for starting the scheduler in the server
// monolith before tasks #5 (LifecycleService) and #14 (ProvisionService) are
// fully implemented.
//
// BuildStubConfig returns a ready-to-use Config whose lifecycle and provision
// dependencies are replaced with no-op stubs. The scheduler's three loops
// still run and hold leader leases, but detect/orchestrate ticks are no-ops.
// Replace the stubs with real implementations once #5/#14 are ready.
import (
"context"
"database/sql"
"github.com/redis/go-redis/v9"
"github.com/wangjia/pangolin/server/internal/nodes"
"github.com/wangjia/pangolin/server/internal/scheduler/detect"
"github.com/wangjia/pangolin/server/internal/scheduler/orchestrate"
"github.com/wangjia/pangolin/server/internal/scheduler/probe"
)
// BuildRealConfig wires the scheduler against the real lifecycle (SQL-backed)
// and provision (#14) services, replacing the no-op stubs of BuildStubConfig.
//
// - detect/orchestrate lifecycle → SQLLifecycle over the nodes table + LoadCache.
// - provision → provisionAdapter over the real provision.Service.
//
// Without vendor credentials the provision service's CreateNode fails (a pending
// replacement simply stays pending), but all lifecycle reads/transitions operate
// on real node rows. Capacity monitoring is left disabled (nil) as in the stub.
//
// e2e behaviour (auto-detect → drain → replace → provision) is verifiable only
// against a real node fleet with vendor provisioning configured.
func BuildRealConfig(
rdb *redis.Client,
probeStore *probe.Store,
db *sql.DB,
loads *nodes.LoadCache,
provSvc provisionSvc,
provStore provisionResolver,
) Config {
sqlLC := NewSQLLifecycle(db, loads)
detectLC := NewDetectLifecycle(sqlLC)
orchLC := NewOrchestrateLifecycle(sqlLC)
prov := NewProvisionAdapter(provSvc, provStore)
streaks := detect.NewStreakStore(rdb)
engine := detect.NewEngine(probeStore, detectLC, streaks, rdb, nil, nil)
replacer := orchestrate.NewReplacer(orchestrate.Config{
RDB: rdb,
Prov: prov,
LC: orchLC,
Snaps: probeStore,
Breaker: nil, // StubBreaker default until 15F policy tuned
Notifier: nil, // LogNotifier default
Clock: nil,
})
grayscale := orchestrate.NewGrayscale(rdb, orchLC, nil)
return Config{
RDB: rdb,
Engine: engine,
Replacer: replacer,
Grayscale: grayscale,
}
}
// BuildStubConfig constructs a scheduler Config wired entirely with in-process
// stubs. Useful for:
// - Starting the scheduler before tasks #5 / #14 are delivered (no-op ticks).
// - Integration tests that need a scheduler goroutine to be running.
//
// To swap in real implementations once they exist:
//
// cfg := scheduler.BuildStubConfig(rdb, probeStore)
// cfg.Engine = detect.NewEngine(realLC, ...)
// cfg.Replacer = orchestrate.NewReplacer(orchestrate.Config{LC: realLC, Prov: realProv, ...})
// cfg.Grayscale = orchestrate.NewGrayscale(rdb, realLC, nil)
// sched := scheduler.New(cfg)
func BuildStubConfig(rdb *redis.Client, probeStore *probe.Store) Config {
// Stub lifecycle for 15D — no nodes registered, so Tick is a safe no-op.
stubDetectLC := detect.NewMockLifecycle(nil)
// Stub lifecycle and provision for 15E.
stubOrchLC := &stubOrchLC{}
stubProv := &stubProvision{}
streaks := detect.NewStreakStore(rdb)
engine := detect.NewEngine(
probeStore, // 15A probe snapshot reader
stubDetectLC, // 15D lifecycle (stub until #5)
streaks,
rdb,
nil, // notifier — LogNotifier used by default
nil, // config — DefaultConfig used
)
replacer := orchestrate.NewReplacer(orchestrate.Config{
RDB: rdb,
Prov: stubProv, // provision stub until #14
LC: stubOrchLC, // lifecycle stub until #5
Snaps: probeStore,
Breaker: nil, // StubBreaker used by default
Notifier: nil, // LogNotifier used by default
Clock: nil, // RealClock used by default
})
grayscale := orchestrate.NewGrayscale(rdb, stubOrchLC, nil)
return Config{
RDB: rdb,
Engine: engine,
Replacer: replacer,
Grayscale: grayscale,
// Capacity, Notifier, ProbeStore, Prober, Targets: nil → stubs/disabled
}
}
// ─────────────────────────────────────────────────────────────────────────────
// Stub implementations for orchestrate.LifecycleService and ProvisionService
//
// These stubs satisfy the interface contracts while doing nothing harmful.
// Replace with the real services (#5 / #14) when those tasks are complete.
// ─────────────────────────────────────────────────────────────────────────────
// stubOrchLC implements orchestrate.LifecycleService as a no-op.
// GetNode always returns nil (node not found); all mutating calls succeed
// silently. This prevents any orchestration work from proceeding until a real
// lifecycle service is wired in.
type stubOrchLC struct{}
func (stubOrchLC) GetNode(_ context.Context, _ string) (*orchestrate.NodeInfo, error) {
return nil, nil // "node not found" → orchestrate skips this replacement
}
func (stubOrchLC) TransitionStatus(_ context.Context, _, _, _ string, _ map[string]any) (int, error) {
return 1, nil
}
func (stubOrchLC) SetWeight(_ context.Context, _ string, _ int) error { return nil }
func (stubOrchLC) BumpVersion(_ context.Context) error { return nil }
func (stubOrchLC) WriteAuditLog(_ context.Context, _, _, _, _ string) error { return nil }
// stubProvision implements orchestrate.ProvisionService as a no-op.
// CreateNode always returns an error so pending replacements stay pending
// rather than silently creating phantom resources.
type stubProvision struct{}
func (stubProvision) CreateNode(_ context.Context, _ orchestrate.NodeSpec, _ string) (string, error) {
return "", context.DeadlineExceeded // signal "not ready" without alarming
}
func (stubProvision) DestroyNode(_ context.Context, _ string) error { return nil }
func (stubProvision) RotateIP(_ context.Context, _ string) (string, error) { return "", nil }
func (stubProvision) ListProviders(_ context.Context, _, _ string) ([]orchestrate.ProviderInfo, error) {
return nil, nil
}