fix(ops-cleanup): 让 UI 数据保留策略真正生效

UI 上 admin 改的数据保留策略(cron + retention 天数)此前只写入 settings 表的
ops_advanced_settings.data_retention,但 OpsCleanupService 启动时只读
cfg.Ops.Cleanup(config.yaml / 环境变量),从未读取 settings 表,导致 UI 配置
完全不生效——cron 实际仍按默认 0 2 * * * 每日跑、retention 30 天。

改动:
- OpsCleanupService 增加 settingRepo 依赖,新增 effective 配置 + Reload 方法。
  Start/Reload 时从 settings.ops_advanced_settings.data_retention 覆盖
  cfg.Ops.Cleanup(Enabled、Schedule、*RetentionDays),无 settings 时整体
  fallback 到 cfg。runScheduled 顶部刷新一次 effective,让 retention 改动当次
  即生效(schedule/enabled 改动需要 Reload 才换 cron)。
- 用 mu + started/stopped 替换 startOnce/stopOnce 以支持 Reload 重建 cron。
- OpsService 增加 CleanupReloader 接口与 SetCleanupReloader setter;
  UpdateOpsAdvancedSettings 写入后调用 Reload。
- wire 通过 setter 注入 cleanup hook,避免构造期循环依赖。
- 新增单测覆盖 overlay 五种情形 + Update 触发 Reload。
This commit is contained in:
erio
2026-05-04 11:36:41 +08:00
parent df722c9a6e
commit c4598aa9b6
6 changed files with 406 additions and 45 deletions

View File

@@ -0,0 +1,196 @@
//go:build unit
package service
import (
"context"
"encoding/json"
"testing"
"github.com/Wei-Shaw/sub2api/internal/config"
)
// makeOverlayService 构造一个没有 cron / db 的 cleanup service仅用来测试 effective overlay。
func makeOverlayService(repo SettingRepository, base config.OpsCleanupConfig) *OpsCleanupService {
cfg := &config.Config{}
cfg.Ops.Cleanup = base
return &OpsCleanupService{
cfg: cfg,
settingRepo: repo,
}
}
func writeAdvancedSettings(t *testing.T, repo *runtimeSettingRepoStub, dr OpsDataRetentionSettings) {
t.Helper()
adv := OpsAdvancedSettings{DataRetention: dr}
raw, err := json.Marshal(adv)
if err != nil {
t.Fatalf("marshal: %v", err)
}
if err := repo.Set(context.Background(), SettingKeyOpsAdvancedSettings, string(raw)); err != nil {
t.Fatalf("set: %v", err)
}
}
func TestComputeEffective_FallbackToCfgWhenSettingsAbsent(t *testing.T) {
repo := newRuntimeSettingRepoStub()
base := config.OpsCleanupConfig{
Enabled: false,
Schedule: "0 2 * * *",
ErrorLogRetentionDays: 30,
MinuteMetricsRetentionDays: 30,
HourlyMetricsRetentionDays: 30,
}
svc := makeOverlayService(repo, base)
svc.computeEffectiveLocked(context.Background())
if svc.effective != base {
t.Fatalf("expected effective == cfg base, got %#v", svc.effective)
}
}
func TestComputeEffective_SettingsOverridesAll(t *testing.T) {
repo := newRuntimeSettingRepoStub()
writeAdvancedSettings(t, repo, OpsDataRetentionSettings{
CleanupEnabled: true,
CleanupSchedule: "0 * * * *",
ErrorLogRetentionDays: 0,
MinuteMetricsRetentionDays: 7,
HourlyMetricsRetentionDays: 14,
})
base := config.OpsCleanupConfig{
Enabled: false,
Schedule: "0 2 * * *",
ErrorLogRetentionDays: 30,
MinuteMetricsRetentionDays: 30,
HourlyMetricsRetentionDays: 30,
}
svc := makeOverlayService(repo, base)
svc.computeEffectiveLocked(context.Background())
want := config.OpsCleanupConfig{
Enabled: true,
Schedule: "0 * * * *",
ErrorLogRetentionDays: 0,
MinuteMetricsRetentionDays: 7,
HourlyMetricsRetentionDays: 14,
}
if svc.effective != want {
t.Fatalf("effective mismatch:\nwant %#v\n got %#v", want, svc.effective)
}
}
func TestComputeEffective_EmptyScheduleFallbackToCfg(t *testing.T) {
repo := newRuntimeSettingRepoStub()
writeAdvancedSettings(t, repo, OpsDataRetentionSettings{
CleanupEnabled: true,
CleanupSchedule: " ", // 空白被 trim 后视为空
ErrorLogRetentionDays: 5,
MinuteMetricsRetentionDays: 5,
HourlyMetricsRetentionDays: 5,
})
base := config.OpsCleanupConfig{
Enabled: false,
Schedule: "0 2 * * *",
ErrorLogRetentionDays: 30,
MinuteMetricsRetentionDays: 30,
HourlyMetricsRetentionDays: 30,
}
svc := makeOverlayService(repo, base)
svc.computeEffectiveLocked(context.Background())
if svc.effective.Schedule != "0 2 * * *" {
t.Fatalf("expected schedule fallback to cfg, got %q", svc.effective.Schedule)
}
if !svc.effective.Enabled {
t.Fatalf("expected enabled=true from settings")
}
if svc.effective.ErrorLogRetentionDays != 5 {
t.Fatalf("expected retention=5 from settings, got %d", svc.effective.ErrorLogRetentionDays)
}
}
func TestComputeEffective_NegativeRetentionFallsBackToCfg(t *testing.T) {
repo := newRuntimeSettingRepoStub()
writeAdvancedSettings(t, repo, OpsDataRetentionSettings{
CleanupEnabled: true,
CleanupSchedule: "0 * * * *",
ErrorLogRetentionDays: -1,
MinuteMetricsRetentionDays: -1,
HourlyMetricsRetentionDays: -1,
})
base := config.OpsCleanupConfig{
Enabled: false,
Schedule: "0 2 * * *",
ErrorLogRetentionDays: 30,
MinuteMetricsRetentionDays: 60,
HourlyMetricsRetentionDays: 90,
}
svc := makeOverlayService(repo, base)
svc.computeEffectiveLocked(context.Background())
if svc.effective.ErrorLogRetentionDays != 30 ||
svc.effective.MinuteMetricsRetentionDays != 60 ||
svc.effective.HourlyMetricsRetentionDays != 90 {
t.Fatalf("expected retention fallback to cfg, got %#v", svc.effective)
}
}
func TestComputeEffective_BadJSONFallsBackToCfg(t *testing.T) {
repo := newRuntimeSettingRepoStub()
if err := repo.Set(context.Background(), SettingKeyOpsAdvancedSettings, "{not json"); err != nil {
t.Fatalf("set: %v", err)
}
base := config.OpsCleanupConfig{
Enabled: true,
Schedule: "0 3 * * *",
ErrorLogRetentionDays: 30,
MinuteMetricsRetentionDays: 30,
HourlyMetricsRetentionDays: 30,
}
svc := makeOverlayService(repo, base)
svc.computeEffectiveLocked(context.Background())
if svc.effective != base {
t.Fatalf("expected fallback to cfg on bad JSON, got %#v", svc.effective)
}
}
// 验证 OpsService.UpdateOpsAdvancedSettings 写入后会调用 cleanupReloader.Reload。
type fakeCleanupReloader struct {
calls int
last context.Context
err error
}
func (f *fakeCleanupReloader) Reload(ctx context.Context) error {
f.calls++
f.last = ctx
return f.err
}
func TestUpdateOpsAdvancedSettings_TriggersReload(t *testing.T) {
repo := newRuntimeSettingRepoStub()
reloader := &fakeCleanupReloader{}
svc := &OpsService{settingRepo: repo}
svc.SetCleanupReloader(reloader)
cfg := defaultOpsAdvancedSettings()
cfg.DataRetention.CleanupEnabled = true
cfg.DataRetention.CleanupSchedule = "0 * * * *"
cfg.DataRetention.ErrorLogRetentionDays = 3
cfg.DataRetention.MinuteMetricsRetentionDays = 3
cfg.DataRetention.HourlyMetricsRetentionDays = 3
if _, err := svc.UpdateOpsAdvancedSettings(context.Background(), cfg); err != nil {
t.Fatalf("update: %v", err)
}
if reloader.calls != 1 {
t.Fatalf("expected reloader.Reload called once, got %d", reloader.calls)
}
}

View File

@@ -3,6 +3,8 @@ package service
import (
"context"
"database/sql"
"encoding/json"
"errors"
"fmt"
"strings"
"sync"
@@ -45,13 +47,18 @@ type OpsCleanupService struct {
redisClient *redis.Client
cfg *config.Config
channelMonitorSvc *ChannelMonitorService
settingRepo SettingRepository
instanceID string
cron *cron.Cron
startOnce sync.Once
stopOnce sync.Once
// mu 守护 cron 实例切换 + effective 配置切换。
// 这里不再用 startOnce/stopOnce是因为 Reload 需要"停旧 cron 重启新 cron"
// 而 Once 一旦触发就无法再次执行;改为 started/stopped 布尔配合 mu。
mu sync.Mutex
cron *cron.Cron
started bool
stopped bool
effective config.OpsCleanupConfig
warnNoRedisOnce sync.Once
}
@@ -62,6 +69,7 @@ func NewOpsCleanupService(
redisClient *redis.Client,
cfg *config.Config,
channelMonitorSvc *ChannelMonitorService,
settingRepo SettingRepository,
) *OpsCleanupService {
return &OpsCleanupService{
opsRepo: opsRepo,
@@ -69,10 +77,13 @@ func NewOpsCleanupService(
redisClient: redisClient,
cfg: cfg,
channelMonitorSvc: channelMonitorSvc,
settingRepo: settingRepo,
instanceID: uuid.NewString(),
}
}
// Start 首次启动 cron 调度。Enabled / Schedule 由 effective 配置决定settings 优先 cfg
// 重复调用幂等。
func (s *OpsCleanupService) Start() {
if s == nil {
return
@@ -80,54 +91,169 @@ func (s *OpsCleanupService) Start() {
if s.cfg != nil && !s.cfg.Ops.Enabled {
return
}
if s.cfg != nil && !s.cfg.Ops.Cleanup.Enabled {
logger.LegacyPrintf("service.ops_cleanup", "[OpsCleanup] not started (disabled)")
return
}
if s.opsRepo == nil || s.db == nil {
logger.LegacyPrintf("service.ops_cleanup", "[OpsCleanup] not started (missing deps)")
return
}
s.startOnce.Do(func() {
schedule := "0 2 * * *"
if s.cfg != nil && strings.TrimSpace(s.cfg.Ops.Cleanup.Schedule) != "" {
schedule = strings.TrimSpace(s.cfg.Ops.Cleanup.Schedule)
}
loc := time.Local
if s.cfg != nil && strings.TrimSpace(s.cfg.Timezone) != "" {
if parsed, err := time.LoadLocation(strings.TrimSpace(s.cfg.Timezone)); err == nil && parsed != nil {
loc = parsed
}
}
c := cron.New(cron.WithParser(opsCleanupCronParser), cron.WithLocation(loc))
_, err := c.AddFunc(schedule, func() { s.runScheduled() })
if err != nil {
logger.LegacyPrintf("service.ops_cleanup", "[OpsCleanup] not started (invalid schedule=%q): %v", schedule, err)
return
}
s.cron = c
s.cron.Start()
logger.LegacyPrintf("service.ops_cleanup", "[OpsCleanup] started (schedule=%q tz=%s)", schedule, loc.String())
})
s.mu.Lock()
defer s.mu.Unlock()
if s.started || s.stopped {
return
}
s.started = true
if err := s.applyScheduleLocked(context.Background()); err != nil {
logger.LegacyPrintf("service.ops_cleanup", "[OpsCleanup] not started: %v", err)
}
}
// Stop 关闭 cron。幂等。
func (s *OpsCleanupService) Stop() {
if s == nil {
return
}
s.stopOnce.Do(func() {
if s.cron != nil {
ctx := s.cron.Stop()
select {
case <-ctx.Done():
case <-time.After(3 * time.Second):
logger.LegacyPrintf("service.ops_cleanup", "[OpsCleanup] cron stop timed out")
}
s.mu.Lock()
defer s.mu.Unlock()
if s.stopped {
return
}
s.stopped = true
s.stopCronLocked()
}
// stopCronLocked 停掉当前 cron 实例(带 3s 超时)。调用方持锁。
func (s *OpsCleanupService) stopCronLocked() {
if s.cron == nil {
return
}
ctx := s.cron.Stop()
select {
case <-ctx.Done():
case <-time.After(3 * time.Second):
logger.LegacyPrintf("service.ops_cleanup", "[OpsCleanup] cron stop timed out")
}
s.cron = nil
}
// applyScheduleLocked 重新计算 effective 配置并按其 schedule 重建 cron。调用方持锁。
// 若 effective.Enabled=false用户在 UI 关闭清理),停旧 cron 后直接返回,不创建新 cron。
func (s *OpsCleanupService) applyScheduleLocked(ctx context.Context) error {
s.computeEffectiveLocked(ctx)
s.stopCronLocked()
if !s.effective.Enabled {
logger.LegacyPrintf("service.ops_cleanup", "[OpsCleanup] cron disabled by settings")
return nil
}
schedule := strings.TrimSpace(s.effective.Schedule)
if schedule == "" {
schedule = "0 2 * * *"
}
loc := time.Local
if s.cfg != nil && strings.TrimSpace(s.cfg.Timezone) != "" {
if parsed, err := time.LoadLocation(strings.TrimSpace(s.cfg.Timezone)); err == nil && parsed != nil {
loc = parsed
}
})
}
c := cron.New(cron.WithParser(opsCleanupCronParser), cron.WithLocation(loc))
if _, err := c.AddFunc(schedule, func() { s.runScheduled() }); err != nil {
return fmt.Errorf("invalid schedule %q: %w", schedule, err)
}
c.Start()
s.cron = c
logger.LegacyPrintf("service.ops_cleanup",
"[OpsCleanup] scheduled (schedule=%q tz=%s retention_days=err:%d/min:%d/hour:%d)",
schedule, loc.String(),
s.effective.ErrorLogRetentionDays,
s.effective.MinuteMetricsRetentionDays,
s.effective.HourlyMetricsRetentionDays,
)
return nil
}
// Reload 重新读取 ops_advanced_settings.data_retention 并按新配置重建 cron。
// 适用于 admin 在 UI 修改清理设置后立即生效schedule / enabled 改动需要 Reload
// retention 改动 runScheduled 顶部也会刷新,下一次触发即生效)。
// 若 service 还未 Start 或已 StopReload 不做任何事。
func (s *OpsCleanupService) Reload(ctx context.Context) error {
if s == nil {
return nil
}
s.mu.Lock()
defer s.mu.Unlock()
if !s.started || s.stopped {
return nil
}
return s.applyScheduleLocked(ctx)
}
// computeEffectiveLocked 计算"生效配置"并写入 s.effective。调用方持锁。
//
// 优先级UI 写入的 settings.ops_advanced_settings.data_retention权威覆盖 cfg.Ops.Cleanup 的副本。
// - Enabledsettings 直接覆盖
// - Schedulesettings 非空时覆盖,否则保留 cfg
// - *RetentionDayssettings >=0 时覆盖(包括 0=TRUNCATE<0 沿用 cfg
//
// 若 settings 表无该 keyErrSettingNotFound或解析失败整体 fallback 到 cfg.Ops.Cleanup。
func (s *OpsCleanupService) computeEffectiveLocked(ctx context.Context) {
base := config.OpsCleanupConfig{}
if s.cfg != nil {
base = s.cfg.Ops.Cleanup
}
defer func() { s.effective = base }()
if s.settingRepo == nil {
return
}
if ctx == nil {
ctx = context.Background()
}
raw, err := s.settingRepo.GetValue(ctx, SettingKeyOpsAdvancedSettings)
if err != nil {
if !errors.Is(err, ErrSettingNotFound) {
logger.LegacyPrintf("service.ops_cleanup",
"[OpsCleanup] read advanced settings failed, using cfg: %v", err)
}
return
}
var adv OpsAdvancedSettings
if err := json.Unmarshal([]byte(raw), &adv); err != nil {
logger.LegacyPrintf("service.ops_cleanup",
"[OpsCleanup] parse advanced settings failed, using cfg: %v", err)
return
}
dr := adv.DataRetention
base.Enabled = dr.CleanupEnabled
if sched := strings.TrimSpace(dr.CleanupSchedule); sched != "" {
base.Schedule = sched
}
if dr.ErrorLogRetentionDays >= 0 {
base.ErrorLogRetentionDays = dr.ErrorLogRetentionDays
}
if dr.MinuteMetricsRetentionDays >= 0 {
base.MinuteMetricsRetentionDays = dr.MinuteMetricsRetentionDays
}
if dr.HourlyMetricsRetentionDays >= 0 {
base.HourlyMetricsRetentionDays = dr.HourlyMetricsRetentionDays
}
}
// snapshotEffective 取一份 effective 副本runCleanupOnce 等读路径使用)。
func (s *OpsCleanupService) snapshotEffective() config.OpsCleanupConfig {
s.mu.Lock()
defer s.mu.Unlock()
return s.effective
}
// refreshEffectiveBeforeRun 在 cron 触发时刷新 effective让 retention 改动当次即生效。
// schedule 改动不影响当次cron 调度由库管理,需要 Reload 才换 schedule
func (s *OpsCleanupService) refreshEffectiveBeforeRun(ctx context.Context) {
s.mu.Lock()
defer s.mu.Unlock()
s.computeEffectiveLocked(ctx)
}
func (s *OpsCleanupService) runScheduled() {
@@ -138,6 +264,9 @@ func (s *OpsCleanupService) runScheduled() {
ctx, cancel := context.WithTimeout(context.Background(), 30*time.Minute)
defer cancel()
// 让 retention 改动当次生效schedule/enabled 改动需要 Reload
s.refreshEffectiveBeforeRun(ctx)
release, ok := s.tryAcquireLeaderLock(ctx)
if !ok {
return
@@ -209,6 +338,8 @@ func (s *OpsCleanupService) runCleanupOnce(ctx context.Context) (opsCleanupDelet
return out, nil
}
effective := s.snapshotEffective()
batchSize := 5000
now := time.Now().UTC()
@@ -223,7 +354,7 @@ func (s *OpsCleanupService) runCleanupOnce(ctx context.Context) (opsCleanupDelet
}
// Error-like tables: error logs / retry attempts / alert events / system logs / cleanup audits.
if cutoff, truncate, ok := opsCleanupPlan(now, s.cfg.Ops.Cleanup.ErrorLogRetentionDays); ok {
if cutoff, truncate, ok := opsCleanupPlan(now, effective.ErrorLogRetentionDays); ok {
n, err := runOne(truncate, cutoff, "ops_error_logs", "created_at", false)
if err != nil {
return out, err
@@ -256,7 +387,7 @@ func (s *OpsCleanupService) runCleanupOnce(ctx context.Context) (opsCleanupDelet
}
// Minute-level metrics snapshots.
if cutoff, truncate, ok := opsCleanupPlan(now, s.cfg.Ops.Cleanup.MinuteMetricsRetentionDays); ok {
if cutoff, truncate, ok := opsCleanupPlan(now, effective.MinuteMetricsRetentionDays); ok {
n, err := runOne(truncate, cutoff, "ops_system_metrics", "created_at", false)
if err != nil {
return out, err
@@ -265,7 +396,7 @@ func (s *OpsCleanupService) runCleanupOnce(ctx context.Context) (opsCleanupDelet
}
// Pre-aggregation tables (hourly/daily).
if cutoff, truncate, ok := opsCleanupPlan(now, s.cfg.Ops.Cleanup.HourlyMetricsRetentionDays); ok {
if cutoff, truncate, ok := opsCleanupPlan(now, effective.HourlyMetricsRetentionDays); ok {
n, err := runOne(truncate, cutoff, "ops_metrics_hourly", "bucket_start", false)
if err != nil {
return out, err

View File

@@ -54,6 +54,24 @@ type OpsService struct {
geminiCompatService *GeminiMessagesCompatService
antigravityGatewayService *AntigravityGatewayService
systemLogSink *OpsSystemLogSink
// cleanupReloader 由 wire 在 OpsCleanupService 构造完成后通过 SetCleanupReloader 注入。
// 解耦避免 OpsService -> OpsCleanupService 的硬依赖cleanup 也读 settings会循环
cleanupReloader CleanupReloader
}
// CleanupReloader 由 OpsCleanupService 实现。
// UpdateOpsAdvancedSettings 写入新配置后调用 Reload让 schedule/enabled 改动立刻生效。
type CleanupReloader interface {
Reload(ctx context.Context) error
}
// SetCleanupReloader 由 wire 注入 cleanup hook构造期循环依赖的解耦点
func (s *OpsService) SetCleanupReloader(r CleanupReloader) {
if s == nil {
return
}
s.cleanupReloader = r
}
func NewOpsService(

View File

@@ -4,6 +4,7 @@ import (
"context"
"encoding/json"
"errors"
"github.com/Wei-Shaw/sub2api/internal/pkg/logger"
"strings"
"time"
)
@@ -477,6 +478,14 @@ func (s *OpsService) UpdateOpsAdvancedSettings(ctx context.Context, cfg *OpsAdva
return nil, err
}
// notify cleanup service to reload schedule/enabled.
if s.cleanupReloader != nil {
if rerr := s.cleanupReloader.Reload(ctx); rerr != nil {
logger.LegacyPrintf("service.ops_settings",
"[OpsSettings] cleanup reload after advanced-settings update failed: %v", rerr)
}
}
updated := &OpsAdvancedSettings{}
_ = json.Unmarshal(raw, updated)
return updated, nil

View File

@@ -271,15 +271,22 @@ func ProvideOpsAlertEvaluatorService(
// ProvideOpsCleanupService creates and starts OpsCleanupService (cron scheduled).
// channelMonitorSvc 让维护任务(聚合 + 历史/聚合软删)跟随 ops 清理 cron 一起跑,
// 共享 leader lock + heartbeat。
// settingRepo 让 cleanup service 自己读 ops_advanced_settings.data_retention 覆盖 cfg
// opsService 用来反向注入 cleanup hook以便 UI 改清理设置时能 Reload cron。
func ProvideOpsCleanupService(
opsRepo OpsRepository,
db *sql.DB,
redisClient *redis.Client,
cfg *config.Config,
channelMonitorSvc *ChannelMonitorService,
settingRepo SettingRepository,
opsService *OpsService,
) *OpsCleanupService {
svc := NewOpsCleanupService(opsRepo, db, redisClient, cfg, channelMonitorSvc)
svc := NewOpsCleanupService(opsRepo, db, redisClient, cfg, channelMonitorSvc, settingRepo)
svc.Start()
if opsService != nil {
opsService.SetCleanupReloader(svc)
}
return svc
}