feat(log): decouple log storage from ClickHouse with switchable logstore

- New internal/repository/logstore abstraction: exported domain interfaces
  (AccessLogStore/ObservabilityStore/UserAccessLogStore/StatusStore),
  config-driven provider (Active/Build/Migrating/SetConfigReader), GORM
  implementation for PostgreSQL/SQLite (incl. hourly rollups computed in
  real time, migration listers, PG partition maintenance), and a ClickHouse
  wrapper preserving the native batch path; repository facade delegates to
  logstore; import-lint test enforces apps never import analyticsrepo.
- ClickHouse is now optional: the log DB is either the main DB (postgres
  when database.enabled, else sqlite) or clickhouse; boot validation +
  first-run seed; log_database / log_db_migration are protected keys.
- New user task 切换日志数据库 (of_log_db_switch): freeze log writes,
  drain batch writers, copy all 6 raw log tables by id (preserving IDs)
  with target-partition pre-creation for PG, flip log_database on success,
  clear the freeze flag on failure.
- Per-store retention (log_retention_days_*) with expiry cleanup folded
  into the daily system_cleanup task; legacy database_auto_cleanup_* and
  of_database_auto_cleanup decommissioned.
- goose migrations: 6 log tables in PG (2 monthly-partitioned) + SQLite,
  retention config seeds, schedule cleanup; GET
  /api/v1/admin/status/log-database endpoint; frontend retention settings,
  switch-task UI and status badge; changelog and docs updated.

docs(plan): log database decoupling implementation plan

docs(design): log database decoupling design (ClickHouse optional)
This commit is contained in:
ryan
2026-08-08 11:36:56 +08:00
parent 734fe45baa
commit 7d71f1e4e1
95 changed files with 10569 additions and 3728 deletions
+125
View File
@@ -0,0 +1,125 @@
// Copyright 2026 Arctel.net
// SPDX-License-Identifier: Apache-2.0
package logstore
import (
"context"
"fmt"
"strconv"
"time"
"github.com/Rain-kl/Wavelet/internal/model"
)
// CleanupSummary 汇总本次清理结果。
type CleanupSummary struct {
ActiveDatabase string `json:"active_database"`
RetentionDays int `json:"retention_days"`
Deleted int64 `json:"deleted"`
// Tables 记录本次清理的物理表简写名(去掉 of_ 前缀,如 node_access_logs 对应
// of_node_access_logs;CH 侧物理表名相同,简写仅便于状态展示)。
Tables []string `json:"tables"`
}
// defaultLogRetentionDays 默认日志保留天数(配置缺失/非法时回退)。
const defaultLogRetentionDays = 90
// partitionLeadMonths 清理时确保「当前月 + 未来 2 个月」分区持续存在。
const partitionLeadMonths = 2
// retentionDaysForActive 按当前激活库读取保留天数(默认 90)。
func retentionDaysForActive(ctx context.Context) int {
key := model.ConfigKeyLogRetentionDaysPostgres
if dbName, _ := resolveDatabase(ctx); dbName == dbNameSQLite {
key = model.ConfigKeyLogRetentionDaysSQLite
} else if dbName == dbNameClickHouse {
key = model.ConfigKeyLogRetentionDaysClickHouse
}
v, err := getConfig(ctx, key)
if err != nil {
return defaultLogRetentionDays
}
days, perr := strconv.Atoi(v)
if perr != nil || days <= 0 {
return defaultLogRetentionDays
}
return days
}
// CleanupExpired 按当前激活库保留天数清理过期日志(每日由 system_cleanup 调用)。
func CleanupExpired(ctx context.Context) (*CleanupSummary, error) {
s, err := Active(ctx)
if err != nil {
return nil, err
}
days := retentionDaysForActive(ctx)
cutoff := time.Now().AddDate(0, 0, -days)
summary := &CleanupSummary{RetentionDays: days, Tables: []string{}}
summary.ActiveDatabase, _ = resolveDatabase(ctx)
// PG 分区表仅在迁移时预建「当前+2 月」分区,此处确保分区持续存在,
// 否则跨月后新写入会报 "no partition of relation found"(SQLite/CH 为 no-op)。
now := time.Now().UTC()
if err := s.AccessLogs.EnsurePartitions(ctx, now, now.AddDate(0, partitionLeadMonths, 0)); err != nil {
return nil, fmt.Errorf("ensure partitions: %w", err)
}
if err := cleanupTable("node_access_logs", func() (int64, error) {
return s.AccessLogs.DeleteBefore(ctx, cutoff)
}, summary); err != nil {
return nil, err
}
if err := cleanupTable("metric_snapshots", func() (int64, error) {
return s.Observability.DeleteMetricSnapshotsBefore(ctx, cutoff)
}, summary); err != nil {
return nil, err
}
if err := cleanupTable("edge_health", func() (int64, error) {
return s.Observability.DeleteEdgeHealthBefore(ctx, cutoff)
}, summary); err != nil {
return nil, err
}
if err := cleanupTable("obs_frps", func() (int64, error) {
return s.Observability.DeleteNodeObservationFrpsBefore(ctx, cutoff)
}, summary); err != nil {
return nil, err
}
if err := cleanupTable("obs_frpc", func() (int64, error) {
return s.Observability.DeleteNodeObservationFrpcBefore(ctx, cutoff)
}, summary); err != nil {
return nil, err
}
return summary, nil
}
func cleanupTable(name string, fn func() (int64, error), summary *CleanupSummary) error {
n, err := fn()
if err != nil {
return fmt.Errorf("cleanup %s: %w", name, err)
}
summary.Deleted += n
summary.Tables = append(summary.Tables, name)
return nil
}
// partitionStatementsRange 生成覆盖 [from, to] 全部月份的两表分区 DDL,
// 幂等 CREATE TABLE IF NOT EXISTS ... PARTITION OF ... FOR VALUES FROM ... TO ...。
// 入参为任意时间点:按各自所在月份生成,含 from 月与 to 月(to 常用 max+1 月兜底)。
func partitionStatementsRange(from, to time.Time) []string {
var out []string
start := time.Date(from.Year(), from.Month(), 1, 0, 0, 0, 0, time.UTC)
end := time.Date(to.Year(), to.Month(), 1, 0, 0, 0, 0, time.UTC).AddDate(0, 1, 0)
for ; start.Before(end); start = start.AddDate(0, 1, 0) {
monthEnd := start.AddDate(0, 1, 0)
suffix := start.Format("200601")
fromDay := start.Format("2006-01-02")
toDay := monthEnd.Format("2006-01-02")
for _, table := range []string{"of_node_access_logs", "w_user_access_logs"} {
out = append(out, fmt.Sprintf(
"CREATE TABLE IF NOT EXISTS %s_%s PARTITION OF %s FOR VALUES FROM ('%s') TO ('%s')",
table, suffix, table, fromDay, toDay))
}
}
return out
}
@@ -0,0 +1,223 @@
// Copyright 2026 Arctel.net
// SPDX-License-Identifier: Apache-2.0
package logstore
import (
"context"
"fmt"
"strings"
"sync/atomic"
"testing"
"time"
"github.com/glebarez/sqlite"
"gorm.io/gorm"
"gorm.io/gorm/logger"
db "github.com/Rain-kl/Wavelet/internal/infra/persistence"
"github.com/Rain-kl/Wavelet/internal/model"
analyticsmodel "github.com/Rain-kl/Wavelet/internal/model/analytics"
)
// cleanupTestModels 清理涉及的 5 张日志/可观测表。
func cleanupTestModels() []any {
return []any{
&analyticsmodel.NodeAccessLog{},
&analyticsmodel.NodeMetricSnapshot{},
&analyticsmodel.NodeEdgeHealth{},
&analyticsmodel.NodeObsFrps{},
&analyticsmodel.NodeObsFrpc{},
}
}
// newCleanupTestDB 构造内存 sqlite 库并注入 db.DB(CleanupExpired 经 Active → buildStore 使用)。
func newCleanupTestDB(t *testing.T) *gorm.DB {
t.Helper()
dsn := fmt.Sprintf("file:logstore-cleanup-%d?mode=memory&cache=shared", atomic.AddInt64(&testGormStoreSeq, 1))
gdb, err := gorm.Open(sqlite.Open(dsn), &gorm.Config{Logger: logger.Default.LogMode(logger.Silent)})
if err != nil {
t.Fatalf("open sqlite: %v", err)
}
if err := gdb.AutoMigrate(cleanupTestModels()...); err != nil {
t.Fatalf("automigrate: %v", err)
}
db.SetDB(gdb)
t.Cleanup(func() { db.SetDB(nil) })
return gdb
}
// TestCleanupExpiredSQLite 验证 sqlite 激活库的过期日志清理:
// 注入 log_retention_days_sqlite=30,40 天前的 5 表记录被删、昨天的保留。
func TestCleanupExpiredSQLite(t *testing.T) {
ResetForTest()
SetConfigReader(func(_ context.Context, key string) (string, error) {
switch key {
case logDatabaseKey:
return "sqlite", nil
case model.ConfigKeyLogRetentionDaysSQLite:
return "30", nil
}
return "", nil
})
defer ResetForTest()
gdb := newCleanupTestDB(t)
ctx := context.Background()
old := time.Now().AddDate(0, 0, -40).UTC()
recent := time.Now().AddDate(0, 0, -1).UTC()
if err := gdb.Create([]analyticsmodel.NodeAccessLog{
{ID: 1, NodeID: "n1", LoggedAt: old, RemoteAddr: "1.1.1.1"},
{ID: 2, NodeID: "n1", LoggedAt: recent, RemoteAddr: "2.2.2.2"},
}).Error; err != nil {
t.Fatalf("seed node access logs: %v", err)
}
if err := gdb.Create([]analyticsmodel.NodeMetricSnapshot{
{ID: 1, NodeID: "n1", CapturedAt: old},
{ID: 2, NodeID: "n1", CapturedAt: recent},
}).Error; err != nil {
t.Fatalf("seed metric snapshots: %v", err)
}
if err := gdb.Create([]analyticsmodel.NodeEdgeHealth{
{ID: 1, NodeID: "n1", CapturedAt: old},
{ID: 2, NodeID: "n1", CapturedAt: recent},
}).Error; err != nil {
t.Fatalf("seed edge health: %v", err)
}
if err := gdb.Create([]analyticsmodel.NodeObsFrps{
{ID: 1, NodeID: "n1", CapturedAt: old},
{ID: 2, NodeID: "n1", CapturedAt: recent},
}).Error; err != nil {
t.Fatalf("seed obs frps: %v", err)
}
if err := gdb.Create([]analyticsmodel.NodeObsFrpc{
{ID: 1, NodeID: "n1", CapturedAt: old},
{ID: 2, NodeID: "n1", CapturedAt: recent},
}).Error; err != nil {
t.Fatalf("seed obs frpc: %v", err)
}
summary, err := CleanupExpired(ctx)
if err != nil {
t.Fatalf("CleanupExpired: %v", err)
}
if summary.ActiveDatabase != "sqlite" {
t.Fatalf("ActiveDatabase = %q, want sqlite", summary.ActiveDatabase)
}
if summary.RetentionDays != 30 {
t.Fatalf("RetentionDays = %d, want 30", summary.RetentionDays)
}
if summary.Deleted != 5 {
t.Fatalf("Deleted = %d, want 5", summary.Deleted)
}
if len(summary.Tables) != 5 {
t.Fatalf("Tables = %v, want 5 tables", summary.Tables)
}
assertCount := func(m any, want int64, label string) {
t.Helper()
var n int64
if err := gdb.Model(m).Count(&n).Error; err != nil {
t.Fatalf("count %s: %v", label, err)
}
if n != want {
t.Fatalf("%s count = %d, want %d", label, n, want)
}
}
assertCount(&analyticsmodel.NodeAccessLog{}, 1, "node_access_logs")
assertCount(&analyticsmodel.NodeMetricSnapshot{}, 1, "metric_snapshots")
assertCount(&analyticsmodel.NodeEdgeHealth{}, 1, "edge_health")
assertCount(&analyticsmodel.NodeObsFrps{}, 1, "obs_frps")
assertCount(&analyticsmodel.NodeObsFrpc{}, 1, "obs_frpc")
var kept analyticsmodel.NodeAccessLog
if err := gdb.First(&kept).Error; err != nil {
t.Fatalf("recent node access log missing: %v", err)
}
if kept.ID != 2 {
t.Fatalf("kept log ID = %d, want 2 (recent)", kept.ID)
}
}
// TestRetentionDaysForActive 覆盖保留天数读取:按激活库选 key、非法值回退默认 90。
func TestRetentionDaysForActive(t *testing.T) {
ResetForTest()
SetConfigReader(func(_ context.Context, key string) (string, error) {
switch key {
case logDatabaseKey:
return "sqlite", nil
case model.ConfigKeyLogRetentionDaysSQLite:
return "30", nil
}
return "", nil
})
if got := retentionDaysForActive(context.Background()); got != 30 {
t.Fatalf("retentionDaysForActive = %d, want 30", got)
}
// 非法值(非数字/<=0)回退默认 90。
SetConfigReader(func(_ context.Context, key string) (string, error) {
switch key {
case logDatabaseKey:
return "postgres", nil
case model.ConfigKeyLogRetentionDaysPostgres:
return "abc", nil
}
return "", nil
})
if got := retentionDaysForActive(context.Background()); got != 90 {
t.Fatalf("retentionDaysForActive invalid value = %d, want 90", got)
}
// reader 报错回退默认 90。
SetConfigReader(func(_ context.Context, _ string) (string, error) {
return "", fmt.Errorf("boom")
})
if got := retentionDaysForActive(context.Background()); got != 90 {
t.Fatalf("retentionDaysForActive reader error = %d, want 90", got)
}
}
// TestPartitionStatements 验证 PG 分区 DDL 生成:当前月 + 未来 2 个月 × 2 表,
// 幂等 PARTITION OF 语句与迁移 SQL 命名一致(含跨年)。
func TestPartitionStatements(t *testing.T) {
now := time.Date(2026, 8, 15, 10, 0, 0, 0, time.UTC)
stmts := partitionStatementsRange(now, now.AddDate(0, 2, 0))
if len(stmts) != 6 {
t.Fatalf("partitionStatements len = %d, want 6", len(stmts))
}
want := []string{
"CREATE TABLE IF NOT EXISTS of_node_access_logs_202608 PARTITION OF of_node_access_logs FOR VALUES FROM ('2026-08-01') TO ('2026-09-01')",
"CREATE TABLE IF NOT EXISTS w_user_access_logs_202608 PARTITION OF w_user_access_logs FOR VALUES FROM ('2026-08-01') TO ('2026-09-01')",
"CREATE TABLE IF NOT EXISTS of_node_access_logs_202609 PARTITION OF of_node_access_logs FOR VALUES FROM ('2026-09-01') TO ('2026-10-01')",
"CREATE TABLE IF NOT EXISTS w_user_access_logs_202609 PARTITION OF w_user_access_logs FOR VALUES FROM ('2026-09-01') TO ('2026-10-01')",
"CREATE TABLE IF NOT EXISTS of_node_access_logs_202610 PARTITION OF of_node_access_logs FOR VALUES FROM ('2026-10-01') TO ('2026-11-01')",
"CREATE TABLE IF NOT EXISTS w_user_access_logs_202610 PARTITION OF w_user_access_logs FOR VALUES FROM ('2026-10-01') TO ('2026-11-01')",
}
for i, w := range want {
if stmts[i] != w {
t.Fatalf("stmt[%d] = %q, want %q", i, stmts[i], w)
}
}
// 跨年:2026-11 → 202611, 202612, 202701。
nov := time.Date(2026, 11, 1, 0, 0, 0, 0, time.UTC)
suffixes := []string{"202611", "202612", "202701"}
for _, stmt := range partitionStatementsRange(nov, nov.AddDate(0, 2, 0)) {
if !hasAnySuffix(stmt, suffixes) {
t.Fatalf("statement lacks expected month suffix: %s", stmt)
}
}
}
func hasAnySuffix(stmt string, suffixes []string) bool {
for _, table := range []string{"of_node_access_logs", "w_user_access_logs"} {
for _, suf := range suffixes {
if strings.Contains(stmt, table+"_"+suf) {
return true
}
}
}
return false
}
@@ -0,0 +1,731 @@
// Copyright 2026 Arctel.net
// SPDX-License-Identifier: Apache-2.0
package logstore
import (
"context"
"errors"
"fmt"
"math"
"time"
"github.com/ClickHouse/clickhouse-go/v2/lib/driver"
db "github.com/Rain-kl/Wavelet/internal/infra/persistence"
"github.com/Rain-kl/Wavelet/internal/model"
analyticsmodel "github.com/Rain-kl/Wavelet/internal/model/analytics"
analyticsrepo "github.com/Rain-kl/Wavelet/internal/repository/analytics"
)
// clickhouseLogStore 实现 AccessLogStore / ObservabilityStore / StatusStore,
// 逐方法委托 analyticsrepo(CH 原生 batch 写入,零性能损耗)。
// UserAccessLogStore 由 clickhouseUserAccessLogStore 实现(List/Count 方法名已被
// AccessLogStore 占用,Go 不允许同名不同签名方法)。
type clickhouseLogStore struct {
// skipFreeze 为 true 时跳过迁移冻结检查(仅迁移目标 store 使用)。
skipFreeze bool
}
func newClickHouseStore() *clickhouseLogStore { return &clickhouseLogStore{} }
// 编译期断言。
var (
_ AccessLogStore = (*clickhouseLogStore)(nil)
_ ObservabilityStore = (*clickhouseLogStore)(nil)
_ StatusStore = (*clickhouseLogStore)(nil)
_ UserAccessLogStore = (*clickhouseUserAccessLogStore)(nil)
)
func chConnErr() error {
if !db.ChConnReady() {
return errors.New("clickhouse connection is not initialized")
}
return nil
}
// ensureWritable 迁移冻结期拒绝写入。
func (s *clickhouseLogStore) ensureWritable(ctx context.Context) error {
if !s.skipFreeze && Migrating(ctx) {
return ErrMigrating
}
return nil
}
// ---- AccessLogStore ----
// InsertBatch 节点访问日志写入入口:冻结检查后经 hook 入队(异步),不直接落库。
func (s *clickhouseLogStore) InsertBatch(ctx context.Context, records []*model.OpenFlareAccessLog) error {
if err := s.ensureWritable(ctx); err != nil {
return err
}
rows := make([]analyticsmodel.NodeAccessLog, 0, len(records))
for _, r := range records {
if r == nil {
continue
}
rows = append(rows, toAnalyticsNodeAccessLog(r))
}
if h := currentAccessLogHooks().QueueNodeAccessLogs; h != nil {
h(rows)
}
return nil
}
// BatchInsertNodeAccessLogs 是 batchwriter flush 目标:CH 原生批量写入。
func (s *clickhouseLogStore) BatchInsertNodeAccessLogs(ctx context.Context, rows []analyticsmodel.NodeAccessLog) error {
if err := s.ensureWritable(ctx); err != nil {
return err
}
return analyticsrepo.BatchInsertNodeAccessLogs(ctx, rows)
}
func (s *clickhouseLogStore) List(ctx context.Context, query model.OpenFlareAccessLogQuery) ([]*model.OpenFlareAccessLog, error) {
rows, err := analyticsrepo.ListNodeAccessLogs(ctx, toNodeAccessLogFilter(query))
if err != nil {
return nil, err
}
return fromAnalyticsNodeAccessLogs(rows), nil
}
func (s *clickhouseLogStore) Count(ctx context.Context, query model.OpenFlareAccessLogQuery) (int64, int64, int64, error) {
return analyticsrepo.CountNodeAccessLogs(ctx, toNodeAccessLogFilter(query))
}
func (s *clickhouseLogStore) RegionCounts(ctx context.Context, nodeID string, since time.Time, limit int) ([]*model.OpenFlareAccessLogRegionCount, error) {
rows, err := analyticsrepo.RegionCountsNodeAccessLogs(ctx, nodeID, since, limit)
if err != nil {
return nil, err
}
out := make([]*model.OpenFlareAccessLogRegionCount, len(rows))
for i, r := range rows {
out[i] = &model.OpenFlareAccessLogRegionCount{Region: r.Region, Count: r.Count}
}
return out, nil
}
func (s *clickhouseLogStore) BucketAggregates(ctx context.Context, query model.OpenFlareAccessLogQuery, bucketSeconds int64) ([]analyticsmodel.NodeAccessLogBucketAggregate, error) {
return analyticsrepo.BucketAggregatesNodeAccessLogs(ctx, toNodeAccessLogFilter(query), bucketSeconds)
}
func (s *clickhouseLogStore) CountBuckets(ctx context.Context, query model.OpenFlareAccessLogQuery, bucketSeconds int64) (int64, error) {
return analyticsrepo.CountBucketAggregatesNodeAccessLogs(ctx, toNodeAccessLogFilter(query), bucketSeconds)
}
func (s *clickhouseLogStore) BucketDimensions(ctx context.Context, query model.OpenFlareAccessLogQuery, column string, bucketSeconds int64) ([]analyticsmodel.NodeAccessLogBucketDimension, error) {
return analyticsrepo.BucketDimensionsNodeAccessLogs(ctx, toNodeAccessLogFilter(query), column, bucketSeconds)
}
func (s *clickhouseLogStore) IPAggregates(ctx context.Context, query model.OpenFlareAccessLogQuery, exactRemoteAddr bool) ([]analyticsmodel.NodeAccessLogIPAggregate, error) {
return analyticsrepo.IPAggregatesNodeAccessLogs(ctx, toNodeAccessLogFilter(query), exactRemoteAddr)
}
func (s *clickhouseLogStore) IPSummaries(ctx context.Context, query model.OpenFlareAccessLogQuery, recentSince time.Time) ([]analyticsmodel.NodeAccessLogIPSummary, error) {
return analyticsrepo.IPSummariesNodeAccessLogs(ctx, toNodeAccessLogFilter(query), recentSince)
}
func (s *clickhouseLogStore) CountIPSummaries(ctx context.Context, query model.OpenFlareAccessLogQuery) (int64, error) {
return analyticsrepo.CountIPSummaryNodeAccessLogs(ctx, toNodeAccessLogFilter(query))
}
func (s *clickhouseLogStore) WAFIPAggregates(ctx context.Context, query model.OpenFlareAccessLogQuery) ([]analyticsmodel.NodeAccessLogWAFIPAggregate, error) {
return analyticsrepo.IPAggregatesForWAFNodeAccessLogs(ctx, toNodeAccessLogFilter(query))
}
func (s *clickhouseLogStore) IPTrend(ctx context.Context, query model.OpenFlareAccessLogQuery, bucketSeconds int64) ([]analyticsmodel.NodeAccessLogIPTrend, error) {
return analyticsrepo.IPTrendNodeAccessLogs(ctx, toNodeAccessLogFilter(query), bucketSeconds)
}
func (s *clickhouseLogStore) TrafficSummary(ctx context.Context, query model.OpenFlareAccessLogQuery) (model.OpenFlareAccessLogTrafficSummary, error) {
row, err := analyticsrepo.TrafficSummaryNodeAccessLogs(ctx, toNodeAccessLogFilter(query))
if err != nil {
return model.OpenFlareAccessLogTrafficSummary{}, err
}
return model.OpenFlareAccessLogTrafficSummary{
RequestCount: row.RequestCount,
ErrorCount: row.ErrorCount,
UniqueIPCount: row.UniqueIPCount,
BytesSent: row.BytesSent,
RequestLength: row.RequestLength,
NodeCount: row.NodeCount,
}, nil
}
func (s *clickhouseLogStore) ValueCounts(ctx context.Context, query model.OpenFlareAccessLogQuery, column string, limit int) ([]model.OpenFlareAccessLogValueCount, error) {
rows, err := analyticsrepo.ValueCountsNodeAccessLogs(ctx, toNodeAccessLogFilter(query), column, limit)
if err != nil {
return nil, err
}
out := make([]model.OpenFlareAccessLogValueCount, len(rows))
for i, r := range rows {
out[i] = model.OpenFlareAccessLogValueCount{Value: r.Value, Count: r.Count}
}
return out, nil
}
func (s *clickhouseLogStore) NodeAggregates(ctx context.Context, query model.OpenFlareAccessLogQuery) ([]model.OpenFlareAccessLogNodeAggregate, error) {
rows, err := analyticsrepo.NodeAggregatesNodeAccessLogs(ctx, toNodeAccessLogFilter(query))
if err != nil {
return nil, err
}
out := make([]model.OpenFlareAccessLogNodeAggregate, len(rows))
for i, r := range rows {
out[i] = model.OpenFlareAccessLogNodeAggregate{NodeID: r.NodeID, RequestCount: r.RequestCount, ErrorCount: r.ErrorCount, UniqueIPCount: r.UniqueIPCount}
}
return out, nil
}
func (s *clickhouseLogStore) DeleteAll(ctx context.Context) (int64, error) {
if err := s.ensureWritable(ctx); err != nil {
return 0, err
}
return analyticsrepo.DeleteAllNodeAccessLogs(ctx)
}
func (s *clickhouseLogStore) DeleteBefore(ctx context.Context, cutoff time.Time) (int64, error) {
if err := s.ensureWritable(ctx); err != nil {
return 0, err
}
return analyticsrepo.DeleteNodeAccessLogsBefore(ctx, cutoff)
}
func (s *clickhouseLogStore) DeleteByNodeBefore(ctx context.Context, nodeID string, before time.Time) (int64, error) {
if err := s.ensureWritable(ctx); err != nil {
return 0, err
}
return analyticsrepo.DeleteNodeAccessLogsByNodeBefore(ctx, nodeID, before)
}
// ListForMigration 按 id 升序分页读取(迁移复制用):直接查询 CH 原生表。
func (s *clickhouseLogStore) ListForMigration(ctx context.Context, afterID uint64, limit int) ([]analyticsmodel.NodeAccessLog, error) {
if err := chConnErr(); err != nil {
return nil, err
}
rows, err := db.ChConn.Query(ctx, `
SELECT `+analyticsmodel.NodeAccessLog{}.InsertColumns()+`
FROM `+analyticsmodel.NodeAccessLog{}.TableName()+`
WHERE id > ?
ORDER BY id ASC
LIMIT ?`, afterID, limitOr(limit, migrationPageSize))
if err != nil {
return nil, fmt.Errorf("list node access logs for migration: %w", err)
}
defer func() { _ = rows.Close() }()
var result []analyticsmodel.NodeAccessLog
for rows.Next() {
var item analyticsmodel.NodeAccessLog
if err := rows.Scan(
&item.ID,
&item.NodeID,
&item.LoggedAt,
&item.RemoteAddr,
&item.Region,
&item.Host,
&item.Path,
&item.UserAgent,
&item.CacheStatus,
&item.StatusCode,
&item.BytesSent,
&item.RequestLength,
&item.RequestTimeMs,
&item.CreatedAt,
); err != nil {
return nil, fmt.Errorf("scan node access log row: %w", err)
}
item.LoggedAt = item.LoggedAt.UTC()
item.CreatedAt = item.CreatedAt.UTC()
result = append(result, item)
}
return result, nil
}
// ---- ObservabilityStore ----
// InsertMetricSnapshot 写入入口:冻结检查 + 经 hook 入队(异步),不直接落库。
func (s *clickhouseLogStore) InsertMetricSnapshot(ctx context.Context, record *model.OpenFlareMetricSnapshot) error {
if record == nil {
return nil
}
if err := s.ensureWritable(ctx); err != nil {
return err
}
if h := currentObservabilityHooks().QueueMetricSnapshot; h != nil {
h(toAnalyticsNodeMetricSnapshot(record))
}
return nil
}
func (s *clickhouseLogStore) ListMetricSnapshots(ctx context.Context, nodeID string, since time.Time, limit int) ([]*model.OpenFlareMetricSnapshot, error) {
rows, err := analyticsrepo.ListNodeMetricSnapshots(ctx, toNodeObservabilityFilter(nodeID, since, limit))
if err != nil {
return nil, err
}
return fromAnalyticsNodeMetricSnapshots(rows), nil
}
func (s *clickhouseLogStore) DeleteAllMetricSnapshots(ctx context.Context) (int64, error) {
if err := s.ensureWritable(ctx); err != nil {
return 0, err
}
return analyticsrepo.DeleteAllNodeMetricSnapshots(ctx)
}
// ListTrafficHourly 委托 analyticsrepo 读 of_access_log_hourly rollup(M5 口径,UV 恒 0)。
func (s *clickhouseLogStore) ListTrafficHourly(ctx context.Context, nodeID string, since time.Time) ([]analyticsmodel.NodeTrafficHourly, error) {
return analyticsrepo.ListNodeTrafficHourly(ctx, toNodeObservabilitySince(nodeID, since))
}
// ListAccessLogHourly 委托 analyticsrepo 读 of_access_log_hourly rollup。
func (s *clickhouseLogStore) ListAccessLogHourly(ctx context.Context, nodeID string, since time.Time) ([]analyticsmodel.AccessLogHourly, error) {
return analyticsrepo.ListAccessLogHourly(ctx, toNodeObservabilitySince(nodeID, since))
}
// ListMetricHourly 委托 analyticsrepo ListNodeMetricHourly:rollup 覆盖窗口时读
// of_node_metric_capacity_hourly,否则按 mergeNodeMetricHourlyPreferRollup 合并 raw 兜底。
func (s *clickhouseLogStore) ListMetricHourly(ctx context.Context, nodeID string, since time.Time) ([]analyticsmodel.NodeMetricHourly, error) {
return analyticsrepo.ListNodeMetricHourly(ctx, toNodeObservabilitySince(nodeID, since))
}
func (s *clickhouseLogStore) DeleteMetricSnapshotsBefore(ctx context.Context, cutoff time.Time) (int64, error) {
if err := s.ensureWritable(ctx); err != nil {
return 0, err
}
return analyticsrepo.DeleteNodeMetricSnapshotsBefore(ctx, cutoff)
}
// BatchInsertNodeMetricSnapshots 是 batchwriter flush 目标:CH 原生批量写入。
func (s *clickhouseLogStore) BatchInsertNodeMetricSnapshots(ctx context.Context, rows []analyticsmodel.NodeMetricSnapshot) error {
if err := s.ensureWritable(ctx); err != nil {
return err
}
return analyticsrepo.BatchInsertNodeMetricSnapshots(ctx, rows)
}
// InsertEdgeHealth 写入入口:冻结检查 + 经 hook 入队(异步),不直接落库。
func (s *clickhouseLogStore) InsertEdgeHealth(ctx context.Context, record *model.OpenFlareEdgeHealth) error {
if record == nil {
return nil
}
if err := s.ensureWritable(ctx); err != nil {
return err
}
if h := currentObservabilityHooks().QueueEdgeHealth; h != nil {
h(toAnalyticsNodeEdgeHealth(record))
}
return nil
}
func (s *clickhouseLogStore) ListEdgeHealth(ctx context.Context, nodeID string, since time.Time, limit int) ([]*model.OpenFlareEdgeHealth, error) {
rows, err := analyticsrepo.ListNodeEdgeHealth(ctx, toNodeObservabilityFilter(nodeID, since, limit))
if err != nil {
return nil, err
}
return fromAnalyticsNodeEdgeHealths(rows), nil
}
func (s *clickhouseLogStore) DeleteAllEdgeHealth(ctx context.Context) (int64, error) {
if err := s.ensureWritable(ctx); err != nil {
return 0, err
}
return analyticsrepo.DeleteAllNodeEdgeHealth(ctx)
}
func (s *clickhouseLogStore) DeleteEdgeHealthBefore(ctx context.Context, cutoff time.Time) (int64, error) {
if err := s.ensureWritable(ctx); err != nil {
return 0, err
}
return analyticsrepo.DeleteNodeEdgeHealthBefore(ctx, cutoff)
}
// BatchInsertNodeEdgeHealth 是 batchwriter flush 目标:CH 原生批量写入。
func (s *clickhouseLogStore) BatchInsertNodeEdgeHealth(ctx context.Context, rows []analyticsmodel.NodeEdgeHealth) error {
if err := s.ensureWritable(ctx); err != nil {
return err
}
return analyticsrepo.BatchInsertNodeEdgeHealth(ctx, rows)
}
// InsertNodeObservationFrps 写入入口:冻结检查 + 经 hook 入队(异步),不直接落库。
func (s *clickhouseLogStore) InsertNodeObservationFrps(ctx context.Context, record *model.OpenFlareNodeObservationFrps) error {
if record == nil {
return nil
}
if err := s.ensureWritable(ctx); err != nil {
return err
}
if h := currentObservabilityHooks().QueueNodeObsFrps; h != nil {
h(toAnalyticsNodeObsFrps(record))
}
return nil
}
func (s *clickhouseLogStore) ListNodeObservationFrps(ctx context.Context, nodeID string, since time.Time, limit int) ([]*model.OpenFlareNodeObservationFrps, error) {
rows, err := analyticsrepo.ListNodeObsFrps(ctx, toNodeObservabilityFilter(nodeID, since, limit))
if err != nil {
return nil, err
}
return fromAnalyticsNodeObsFrps(rows), nil
}
func (s *clickhouseLogStore) DeleteAllNodeObservationFrps(ctx context.Context) (int64, error) {
if err := s.ensureWritable(ctx); err != nil {
return 0, err
}
return analyticsrepo.DeleteAllNodeObsFrps(ctx)
}
func (s *clickhouseLogStore) DeleteNodeObservationFrpsBefore(ctx context.Context, cutoff time.Time) (int64, error) {
if err := s.ensureWritable(ctx); err != nil {
return 0, err
}
return analyticsrepo.DeleteNodeObsFrpsBefore(ctx, cutoff)
}
// BatchInsertNodeObsFrps 是 batchwriter flush 目标:CH 原生批量写入。
func (s *clickhouseLogStore) BatchInsertNodeObsFrps(ctx context.Context, rows []analyticsmodel.NodeObsFrps) error {
if err := s.ensureWritable(ctx); err != nil {
return err
}
return analyticsrepo.BatchInsertNodeObsFrps(ctx, rows)
}
// InsertNodeObservationFrpc 写入入口:冻结检查 + 经 hook 入队(异步),不直接落库。
func (s *clickhouseLogStore) InsertNodeObservationFrpc(ctx context.Context, record *model.OpenFlareNodeObservationFrpc) error {
if record == nil {
return nil
}
if err := s.ensureWritable(ctx); err != nil {
return err
}
if h := currentObservabilityHooks().QueueNodeObsFrpc; h != nil {
h(toAnalyticsNodeObsFrpc(record))
}
return nil
}
func (s *clickhouseLogStore) ListNodeObservationFrpc(ctx context.Context, nodeID string, since time.Time, limit int) ([]*model.OpenFlareNodeObservationFrpc, error) {
rows, err := analyticsrepo.ListNodeObsFrpc(ctx, toNodeObservabilityFilter(nodeID, since, limit))
if err != nil {
return nil, err
}
return fromAnalyticsNodeObsFrpc(rows), nil
}
func (s *clickhouseLogStore) DeleteAllNodeObservationFrpc(ctx context.Context) (int64, error) {
if err := s.ensureWritable(ctx); err != nil {
return 0, err
}
return analyticsrepo.DeleteAllNodeObsFrpc(ctx)
}
func (s *clickhouseLogStore) DeleteNodeObservationFrpcBefore(ctx context.Context, cutoff time.Time) (int64, error) {
if err := s.ensureWritable(ctx); err != nil {
return 0, err
}
return analyticsrepo.DeleteNodeObsFrpcBefore(ctx, cutoff)
}
// MigrationRange 返回 of_node_access_logs.logged_at 的最小/最大值(空表返回零值)。
func (s *clickhouseLogStore) MigrationRange(ctx context.Context) (time.Time, time.Time, error) {
return chMigrationRange(ctx, analyticsmodel.NodeAccessLog{}.TableName(), "logged_at")
}
// EnsurePartitions 是 CH 分支 no-op(CH 无 PG 式分区)。
func (s *clickhouseLogStore) EnsurePartitions(_ context.Context, _, _ time.Time) error {
return nil
}
// chMigrationRange 查询 CH 表时间列 MIN/MAX;空表(NULL)返回零值。
func chMigrationRange(ctx context.Context, table, column string) (time.Time, time.Time, error) {
if err := chConnErr(); err != nil {
return time.Time{}, time.Time{}, err
}
var minTime, maxTime *time.Time
if err := db.ChConn.QueryRow(ctx,
"SELECT min("+column+"), max("+column+") FROM "+table,
).Scan(&minTime, &maxTime); err != nil {
return time.Time{}, time.Time{}, fmt.Errorf("query migration range %s: %w", table, err)
}
if minTime == nil || maxTime == nil {
return time.Time{}, time.Time{}, nil
}
return minTime.UTC(), maxTime.UTC(), nil
}
// BatchInsertNodeObsFrpc 是 batchwriter flush 目标:CH 原生批量写入。
func (s *clickhouseLogStore) BatchInsertNodeObsFrpc(ctx context.Context, rows []analyticsmodel.NodeObsFrpc) error {
if err := s.ensureWritable(ctx); err != nil {
return err
}
return analyticsrepo.BatchInsertNodeObsFrpc(ctx, rows)
}
// ListMetricSnapshotsForMigration 按 id 升序分页读取(迁移复制用)。
func (s *clickhouseLogStore) ListMetricSnapshotsForMigration(ctx context.Context, afterID uint64, limit int) ([]analyticsmodel.NodeMetricSnapshot, error) {
return chListForMigration(ctx, afterID, limit,
analyticsmodel.NodeMetricSnapshot{}.TableName(),
analyticsmodel.NodeMetricSnapshot{}.InsertColumns(),
func(rows driver.Rows) ([]analyticsmodel.NodeMetricSnapshot, error) {
var result []analyticsmodel.NodeMetricSnapshot
for rows.Next() {
var item analyticsmodel.NodeMetricSnapshot
if err := rows.Scan(
&item.ID,
&item.NodeID,
&item.CapturedAt,
&item.CPUUsagePercent,
&item.MemoryUsedBytes,
&item.MemoryTotalBytes,
&item.StorageUsedBytes,
&item.StorageTotalBytes,
&item.DiskReadBytes,
&item.DiskWriteBytes,
&item.NetworkRxBytes,
&item.NetworkTxBytes,
&item.CreatedAt,
); err != nil {
return nil, fmt.Errorf("scan node metric snapshot row: %w", err)
}
item.CapturedAt = item.CapturedAt.UTC()
item.CreatedAt = item.CreatedAt.UTC()
result = append(result, item)
}
return result, nil
})
}
// chObsRow 迁移读取共用的双字段观测行(字符串状态 + 数值计数):
// edge_health(status/connections)与 obs_frpc(tunnel_status/connected_relays_count)同形状。
type chObsRow struct {
ID uint64
NodeID string
CapturedAt time.Time
Status string
Count int64
CreatedAt time.Time
}
// countToInt32 将观测计数转为 int32(防御溢出;观测计数远小于 int32 上限)。
func countToInt32(v int64) int32 {
if v > math.MaxInt32 {
return math.MaxInt32
}
if v < math.MinInt32 {
return math.MinInt32
}
return int32(v)
}
// scanChObsRow 扫描 chObsRow(含 UTC 归一化)。
func scanChObsRow(rows driver.Rows) ([]chObsRow, error) {
var result []chObsRow
for rows.Next() {
var item chObsRow
if err := rows.Scan(
&item.ID,
&item.NodeID,
&item.CapturedAt,
&item.Status,
&item.Count,
&item.CreatedAt,
); err != nil {
return nil, fmt.Errorf("scan observation row: %w", err)
}
item.CapturedAt = item.CapturedAt.UTC()
item.CreatedAt = item.CreatedAt.UTC()
result = append(result, item)
}
return result, nil
}
// ListEdgeHealthForMigration 按 id 升序分页读取(迁移复制用)。
func (s *clickhouseLogStore) ListEdgeHealthForMigration(ctx context.Context, afterID uint64, limit int) ([]analyticsmodel.NodeEdgeHealth, error) {
rows, err := chListForMigration(ctx, afterID, limit,
analyticsmodel.NodeEdgeHealth{}.TableName(),
analyticsmodel.NodeEdgeHealth{}.InsertColumns(),
scanChObsRow)
if err != nil {
return nil, err
}
out := make([]analyticsmodel.NodeEdgeHealth, len(rows))
for i, r := range rows {
out[i] = analyticsmodel.NodeEdgeHealth{ID: r.ID, NodeID: r.NodeID, CapturedAt: r.CapturedAt, Status: r.Status, Connections: r.Count, CreatedAt: r.CreatedAt}
}
return out, nil
}
// ListNodeObsFrpsForMigration 按 id 升序分页读取(迁移复制用)。
func (s *clickhouseLogStore) ListNodeObsFrpsForMigration(ctx context.Context, afterID uint64, limit int) ([]analyticsmodel.NodeObsFrps, error) {
return chListForMigration(ctx, afterID, limit,
analyticsmodel.NodeObsFrps{}.TableName(),
analyticsmodel.NodeObsFrps{}.InsertColumns(),
func(rows driver.Rows) ([]analyticsmodel.NodeObsFrps, error) {
var result []analyticsmodel.NodeObsFrps
for rows.Next() {
var item analyticsmodel.NodeObsFrps
if err := rows.Scan(
&item.ID,
&item.NodeID,
&item.CapturedAt,
&item.FrpsConnections,
&item.FrpsProxyCount,
&item.FrpsClientCount,
&item.FrpsProxies,
&item.CreatedAt,
); err != nil {
return nil, fmt.Errorf("scan node frps observation row: %w", err)
}
item.CapturedAt = item.CapturedAt.UTC()
item.CreatedAt = item.CreatedAt.UTC()
result = append(result, item)
}
return result, nil
})
}
// ListNodeObsFrpcForMigration 按 id 升序分页读取(迁移复制用)。
func (s *clickhouseLogStore) ListNodeObsFrpcForMigration(ctx context.Context, afterID uint64, limit int) ([]analyticsmodel.NodeObsFrpc, error) {
rows, err := chListForMigration(ctx, afterID, limit,
analyticsmodel.NodeObsFrpc{}.TableName(),
analyticsmodel.NodeObsFrpc{}.InsertColumns(),
scanChObsRow)
if err != nil {
return nil, err
}
out := make([]analyticsmodel.NodeObsFrpc, len(rows))
for i, r := range rows {
out[i] = analyticsmodel.NodeObsFrpc{ID: r.ID, NodeID: r.NodeID, CapturedAt: r.CapturedAt, TunnelStatus: r.Status, ConnectedRelaysCount: countToInt32(r.Count), CreatedAt: r.CreatedAt}
}
return out, nil
}
// chListForMigration 执行按 id 升序分页的 CH 原生表查询,并交给 scanner 扫描。
func chListForMigration[T any](ctx context.Context, afterID uint64, limit int, table, columns string, scanner func(driver.Rows) ([]T, error)) ([]T, error) {
if err := chConnErr(); err != nil {
return nil, err
}
rows, err := db.ChConn.Query(ctx, `
SELECT `+columns+`
FROM `+table+`
WHERE id > ?
ORDER BY id ASC
LIMIT ?`, afterID, limitOr(limit, migrationPageSize))
if err != nil {
return nil, fmt.Errorf("list %s for migration: %w", table, err)
}
defer func() { _ = rows.Close() }()
return scanner(rows)
}
// ---- StatusStore ----
// ActiveDatabase 返回当前日志主库名(CH 分支固定 clickhouse)。
func (s *clickhouseLogStore) ActiveDatabase(_ context.Context) (string, error) {
return dbNameClickHouse, nil
}
// ClickHouseOperationalStats 委托 analyticsrepo 汇总 CH 运行状态。
func (s *clickhouseLogStore) ClickHouseOperationalStats(ctx context.Context) (*analyticsmodel.ClickHouseOperationalStats, error) {
return analyticsrepo.GetClickHouseOperationalStats(ctx)
}
// ---- UserAccessLogStore ----
// clickhouseUserAccessLogStore 实现 UserAccessLogStore。clickhouseLogStore 已占用
// List/Count 方法名(AccessLogStore 接口),Go 不允许同名不同签名方法,故用户访问日志
// 用独立类型嵌入同一 clickhouseLogStore(与 userAccessLogGormStore 同构),复用 ensureWritable。
type clickhouseUserAccessLogStore struct {
*clickhouseLogStore
}
func newClickHouseUserAccessLogStore() *clickhouseUserAccessLogStore {
return &clickhouseUserAccessLogStore{clickhouseLogStore: newClickHouseStore()}
}
// BatchInsert 是 batchwriter flush 目标:CH 原生批量写入;冻结期拒绝写入,空批次直接返回。
func (s *clickhouseUserAccessLogStore) BatchInsert(ctx context.Context, logs []analyticsmodel.UserAccessLog) error {
if len(logs) == 0 {
return nil
}
if err := s.ensureWritable(ctx); err != nil {
return err
}
return analyticsrepo.BatchInsert(ctx, logs)
}
// DeleteAll 清空全部用户访问日志(TRUNCATE 语义,迁移「覆盖目标库已有日志」幂等前提用)。
func (s *clickhouseUserAccessLogStore) DeleteAll(ctx context.Context) (int64, error) {
if err := s.ensureWritable(ctx); err != nil {
return 0, err
}
return analyticsrepo.DeleteAllUserAccessLogs(ctx)
}
// ListForMigration 按 id 升序分页读取(迁移复制用)。
func (s *clickhouseUserAccessLogStore) ListForMigration(ctx context.Context, afterID uint64, limit int) ([]analyticsmodel.UserAccessLog, error) {
return chListForMigration(ctx, afterID, limit,
analyticsmodel.UserAccessLog{}.TableName(),
analyticsmodel.UserAccessLog{}.InsertColumns(),
func(rows driver.Rows) ([]analyticsmodel.UserAccessLog, error) {
var result []analyticsmodel.UserAccessLog
for rows.Next() {
var item analyticsmodel.UserAccessLog
if err := rows.Scan(
&item.ID,
&item.UserID,
&item.Path,
&item.Method,
&item.IP,
&item.UserAgent,
&item.Headers,
&item.Status,
&item.Latency,
&item.CreatedAt,
); err != nil {
return nil, fmt.Errorf("scan user access log row: %w", err)
}
item.CreatedAt = item.CreatedAt.UTC()
result = append(result, item)
}
return result, nil
})
}
// MigrationRange 返回 w_user_access_logs.created_at 的最小/最大值(空表返回零值)。
func (s *clickhouseUserAccessLogStore) MigrationRange(ctx context.Context) (time.Time, time.Time, error) {
return chMigrationRange(ctx, analyticsmodel.UserAccessLog{}.TableName(), "created_at")
}
func (s *clickhouseUserAccessLogStore) Count(ctx context.Context, filter analyticsmodel.AccessLogFilter) (uint64, error) {
return analyticsrepo.CountAccessLogs(ctx, filter)
}
func (s *clickhouseUserAccessLogStore) List(ctx context.Context, filter analyticsmodel.AccessLogFilter, page, pageSize int) ([]analyticsmodel.UserAccessLog, uint64, error) {
return analyticsrepo.ListAccessLogs(ctx, filter, page, pageSize)
}
func (s *clickhouseUserAccessLogStore) GetDailyTrend(ctx context.Context, days int) ([]analyticsmodel.DailyTrend, error) {
return analyticsrepo.GetDailyTrend(ctx, days)
}
func (s *clickhouseUserAccessLogStore) GetBrowserDistribution(ctx context.Context, startTime time.Time) ([]analyticsmodel.BrowserShare, error) {
return analyticsrepo.GetBrowserDistribution(ctx, startTime)
}
func (s *clickhouseUserAccessLogStore) GetTopActiveUsers(ctx context.Context, startTime time.Time, limit int) ([]analyticsmodel.TopUser, error) {
return analyticsrepo.GetTopActiveUsers(ctx, startTime, limit)
}
// toNodeObservabilityFilter 构造 CH 可观测查询过滤器(limit<=0 表示不限制)。
func toNodeObservabilityFilter(nodeID string, since time.Time, limit int) analyticsmodel.NodeObservabilityFilter {
return analyticsmodel.NodeObservabilityFilter{
NodeID: nodeID,
Since: since,
Limit: limit,
}
}
// toNodeObservabilitySince 构造不带 limit 的可观测查询过滤器
// (小时级聚合读无需分页,避免传无意义的 0)。
func toNodeObservabilitySince(nodeID string, since time.Time) analyticsmodel.NodeObservabilityFilter {
return analyticsmodel.NodeObservabilityFilter{NodeID: nodeID, Since: since}
}
@@ -0,0 +1,40 @@
// Copyright 2026 Arctel.net
// SPDX-License-Identifier: Apache-2.0
package logstore
import (
"context"
"strings"
"testing"
"time"
db "github.com/Rain-kl/Wavelet/internal/infra/persistence"
)
// TestClickHouseHourlyDelegationRegression 验证 CH 后端小时级聚合读委托 analyticsrepo:
// 未初始化 CH 连接时返回 analyticsrepo 的 "clickhouse connection is not initialized" 错误
// (而非未实现/panic),证明 3 个方法都路由到 CH 原生查询。
func TestClickHouseHourlyDelegationRegression(t *testing.T) {
if db.ChConnReady() {
t.Skip("clickhouse connection initialized; skipping delegation regression")
}
s := newClickHouseStore()
ctx := context.Background()
now := time.Now()
check := func(name string, err error) {
t.Helper()
if err == nil {
t.Fatalf("%s: want clickhouse-not-initialized error, got nil", name)
}
if !strings.Contains(err.Error(), "clickhouse connection is not initialized") {
t.Fatalf("%s: unexpected error %v", name, err)
}
}
_, err := s.ListTrafficHourly(ctx, "n1", now)
check("ListTrafficHourly", err)
_, err = s.ListAccessLogHourly(ctx, "n1", now)
check("ListAccessLogHourly", err)
_, err = s.ListMetricHourly(ctx, "n1", now)
check("ListMetricHourly", err)
}
@@ -0,0 +1,33 @@
// Copyright 2026 Arctel.net
// SPDX-License-Identifier: Apache-2.0
package logstore
import (
"strconv"
)
// timeBucketSQLPostgres 返回 PG 时间分桶表达式(epoch 秒 -> 分桶起点,int64)。
func timeBucketSQLPostgres(column string, bucketSeconds int64) string {
return "(floor(extract(epoch from " + column + ")/" + strconv.FormatInt(bucketSeconds, 10) + ")*" + strconv.FormatInt(bucketSeconds, 10) + ")::bigint"
}
// dailyTrendDateSQLPostgres 返回 PG 按日聚合的日期表达式。
func dailyTrendDateSQLPostgres() string {
return "to_char(created_at, 'YYYY-MM-DD')"
}
// epochSQLPostgres 返回 PG epoch 秒表达式(int64)。
func epochSQLPostgres(column string) string {
return "extract(epoch from " + column + ")::bigint"
}
// textCastSQLPostgres 返回 PG 数值列转文本表达式。
func textCastSQLPostgres(column string) string {
return column + "::text"
}
// distinctNonEmptyCountSQLPostgres 返回 PG 排除空串的 distinct 计数表达式。
func distinctNonEmptyCountSQLPostgres(column string) string {
return "COUNT(DISTINCT " + column + ") FILTER (WHERE " + column + " <> '')"
}
@@ -0,0 +1,85 @@
// Copyright 2026 Arctel.net
// SPDX-License-Identifier: Apache-2.0
package logstore
import (
"strconv"
"gorm.io/gorm"
)
// isPostgresDialect 判断 gorm 句柄是否为 PostgreSQL 方言(否则按 SQLite 处理)。
// Dialector 经 gorm.Config 内嵌提升,Name() 可直接在 DB 上调用。
func isPostgresDialect(db *gorm.DB) bool {
return db != nil && db.Dialector != nil && db.Name() == "postgres"
}
// timeBucketSQLSQLite 返回 SQLite 时间分桶表达式(epoch 秒 -> 分桶起点)。
func timeBucketSQLSQLite(column string, bucketSeconds int64) string {
return "(floor(unixepoch(" + column + ")/" + strconv.FormatInt(bucketSeconds, 10) + ")*" + strconv.FormatInt(bucketSeconds, 10) + ")"
}
// dailyTrendDateSQLSQLite 返回 SQLite 按日聚合的日期表达式。
func dailyTrendDateSQLSQLite() string {
return "strftime('%Y-%m-%d', created_at)"
}
// epochSQLSQLite 返回 SQLite epoch 秒表达式(unixepoch 整数秒)。
func epochSQLSQLite(column string) string {
return "unixepoch(" + column + ")"
}
// textCastSQLSQLite 返回 SQLite 数值列转文本表达式。
func textCastSQLSQLite(column string) string {
return "CAST(" + column + " AS TEXT)"
}
// distinctNonEmptyCountSQLSQLite 返回 SQLite 排除空串的 distinct 计数表达式
// (SQLite 无 FILTER 语法,用 CASE 等价实现)。
func distinctNonEmptyCountSQLSQLite(column string) string {
return "COUNT(DISTINCT CASE WHEN " + column + " <> '' THEN " + column + " END)"
}
// distinctNonEmptyCountSQL 按当前方言返回排除空串的 distinct 计数表达式
// (运行时按 Dialector 分发,默认 SQLite)。
func distinctNonEmptyCountSQL(db *gorm.DB, column string) string {
if isPostgresDialect(db) {
return distinctNonEmptyCountSQLPostgres(column)
}
return distinctNonEmptyCountSQLSQLite(column)
}
// dailyTrendDateSQL 按当前方言返回按日聚合的日期表达式(运行时按 Dialector 分发,默认 SQLite)。
func dailyTrendDateSQL(db *gorm.DB) string {
if isPostgresDialect(db) {
return dailyTrendDateSQLPostgres()
}
return dailyTrendDateSQLSQLite()
}
// epochSQL 按当前方言返回 epoch 秒表达式(运行时按 Dialector 分发,默认 SQLite)。
func epochSQL(db *gorm.DB, column string) string {
if isPostgresDialect(db) {
return epochSQLPostgres(column)
}
return epochSQLSQLite(column)
}
// textCastSQL 按当前方言返回数值列转文本表达式(运行时按 Dialector 分发,默认 SQLite)。
func textCastSQL(db *gorm.DB, column string) string {
if isPostgresDialect(db) {
return textCastSQLPostgres(column)
}
return textCastSQLSQLite(column)
}
// timeBucketSQL 按当前方言返回时间分桶表达式。
// brief 将 PG/SQLite 两版写为同名函数,同包无法共存;log_database 为运行时配置,
// 不能使用编译期 build tag,故按 db.Dialector.Name() 运行时分发(默认 SQLite)。
func timeBucketSQL(db *gorm.DB, column string, bucketSeconds int64) string {
if isPostgresDialect(db) {
return timeBucketSQLPostgres(column, bucketSeconds)
}
return timeBucketSQLSQLite(column, bucketSeconds)
}
+57
View File
@@ -0,0 +1,57 @@
// Copyright 2026 Arctel.net
// SPDX-License-Identifier: Apache-2.0
package logstore
import (
"sync"
analyticsmodel "github.com/Rain-kl/Wavelet/internal/model/analytics"
)
// AccessLogHooks 节点访问日志异步入队回调(由 chwriter 装配)。
type AccessLogHooks struct {
QueueNodeAccessLogs func(logs []analyticsmodel.NodeAccessLog)
}
// ObservabilityHooks 可观测异步入队回调(由 chwriter 装配)。
type ObservabilityHooks struct {
QueueMetricSnapshot func(record analyticsmodel.NodeMetricSnapshot)
QueueEdgeHealth func(record analyticsmodel.NodeEdgeHealth)
QueueNodeObsFrps func(record analyticsmodel.NodeObsFrps)
QueueNodeObsFrpc func(record analyticsmodel.NodeObsFrpc)
}
var (
hooksMu sync.RWMutex
accessLogHooks AccessLogHooks
observabilityHooks ObservabilityHooks
)
// SetAccessLogHooks 注册节点访问日志异步入队回调。
func SetAccessLogHooks(h AccessLogHooks) {
hooksMu.Lock()
accessLogHooks = h
hooksMu.Unlock()
}
// SetObservabilityHooks 注册可观测异步入队回调。
func SetObservabilityHooks(h ObservabilityHooks) {
hooksMu.Lock()
observabilityHooks = h
hooksMu.Unlock()
}
// currentAccessLogHooks 返回当前 hooks 快照(未注册时为 zero value,调用方判空跳过)。
func currentAccessLogHooks() AccessLogHooks {
hooksMu.RLock()
defer hooksMu.RUnlock()
return accessLogHooks
}
// currentObservabilityHooks 返回当前 hooks 快照(未注册时为 zero value,调用方判空跳过)。
func currentObservabilityHooks() ObservabilityHooks {
hooksMu.RLock()
defer hooksMu.RUnlock()
return observabilityHooks
}
@@ -0,0 +1,67 @@
// Copyright 2026 Arctel.net
// SPDX-License-Identifier: Apache-2.0
package logstore
import (
"os/exec"
"strings"
"testing"
)
// forbiddenImports 上层应用禁止直接触碰的底层日志实现。
var forbiddenImports = []string{
"github.com/Rain-kl/Wavelet/internal/repository/analytics",
}
// allowedAnalyticsDelegation 允许直接依赖 analyticsrepo 的委托层:
// - internal/repository:持久化门面,ListOpenFlareLatestMetricSnapshotsSince 的
// CH 快速路径仍直连 analyticsrepo(LIMIT 1 BY node_id);小时级聚合读已改走 logstore;
// - internal/repository/logstore:CH 后端实现按设计委托 analyticsrepo。
//
// 除此之外,依赖闭包内任何包都禁止引入 analyticsrepo。
var allowedAnalyticsDelegation = map[string]bool{
"github.com/Rain-kl/Wavelet/internal/repository": true,
"github.com/Rain-kl/Wavelet/internal/repository/logstore": true,
}
// allowedInfraPersistence 允许 apps 引入的 infra/persistence 子包。
var allowedInfraPersistence = []string{
"github.com/Rain-kl/Wavelet/internal/infra/persistence/batchwriter", // batchwriter 统计类型
"github.com/Rain-kl/Wavelet/internal/infra/persistence/idgen", // 雪花 ID 生成(无日志依赖)
}
func TestAppsMustNotImportLogBackendDirectly(t *testing.T) {
t.Chdir("../../..") // module root,保证 ./internal/apps/... 可解析
// -test 同时列出测试二进制依赖,覆盖仅测试文件引入的底层日志实现。
out, err := exec.Command("go", "list", "-test", "-deps", "-f", `{{.ImportPath}} {{join .Imports " "}}`, "./internal/apps/...").Output()
if err != nil {
t.Fatalf("go list: %v", err)
}
for _, line := range strings.Split(string(out), "\n") {
fields := strings.Fields(line)
if len(fields) == 0 {
continue
}
pkg := fields[0]
for _, forbidden := range forbiddenImports {
for _, imp := range fields[1:] {
if imp == forbidden && !allowedAnalyticsDelegation[pkg] {
t.Errorf("internal/apps must not import %s (via %s)", forbidden, pkg)
}
}
}
if strings.HasPrefix(pkg, "github.com/Rain-kl/Wavelet/internal/infra/persistence/") {
allowed := false
for _, a := range allowedInfraPersistence {
if pkg == a || strings.HasPrefix(pkg, a+"/") {
allowed = true
break
}
}
if !allowed {
t.Errorf("internal/apps must not import infra/persistence subpackage: %s", pkg)
}
}
}
}
@@ -0,0 +1,65 @@
// Copyright 2026 Arctel.net
// SPDX-License-Identifier: Apache-2.0
package logstore
import (
"context"
"testing"
"time"
"github.com/Rain-kl/Wavelet/internal/model"
analyticsmodel "github.com/Rain-kl/Wavelet/internal/model/analytics"
)
// TestGormAccessLogInsertBatchHooks 覆盖访问日志写入入口:
// 冻结检查、hook 入队、不直接落库、flush 后可见(行为与旧 repository clickhouse 包装一致)。
func TestGormAccessLogInsertBatchHooks(t *testing.T) {
ResetForTest()
SetConfigReader(func(_ context.Context, key string) (string, error) {
return "", nil
})
defer ResetForTest()
s := newTestGormStore(t)
ctx := context.Background()
now := time.Now().UTC()
var hooked []analyticsmodel.NodeAccessLog
SetAccessLogHooks(AccessLogHooks{
QueueNodeAccessLogs: func(logs []analyticsmodel.NodeAccessLog) {
hooked = append(hooked, logs...)
},
})
defer SetAccessLogHooks(AccessLogHooks{})
records := []*model.OpenFlareAccessLog{
{NodeID: "n1", LoggedAt: now, RemoteAddr: "1.1.1.1", StatusCode: 200, BytesSent: 100},
{NodeID: "n1", LoggedAt: now, RemoteAddr: "2.2.2.2", StatusCode: 404},
}
if err := s.InsertBatch(ctx, records); err != nil {
t.Fatalf("insert batch: %v", err)
}
if len(hooked) != 2 || hooked[0].RemoteAddr != "1.1.1.1" || hooked[0].BytesSent != 100 || hooked[1].StatusCode != 404 {
t.Fatalf("hook rows mismatch: %+v", hooked)
}
// 写入入口只入队、不直接落库。
rows, err := s.List(ctx, model.OpenFlareAccessLogQuery{NodeID: "n1"})
if err != nil {
t.Fatalf("list: %v", err)
}
if len(rows) != 0 {
t.Fatalf("entry insert must not write rows, got %d", len(rows))
}
// flush 后可见。
if err := s.BatchInsertNodeAccessLogs(ctx, hooked); err != nil {
t.Fatalf("flush: %v", err)
}
rows, err = s.List(ctx, model.OpenFlareAccessLogQuery{NodeID: "n1"})
if err != nil {
t.Fatalf("list after flush: %v", err)
}
if len(rows) != 2 {
t.Fatalf("list after flush want 2, got %d", len(rows))
}
}
+122
View File
@@ -0,0 +1,122 @@
// Copyright 2026 Arctel.net
// SPDX-License-Identifier: Apache-2.0
// Package logstore 提供日志/分析存储抽象:上层只面向本包接口,
// 禁止直接 import internal/repository/analytics 或触碰 db.ChConn/db.ChDB。
package logstore
import (
"context"
"errors"
"time"
"github.com/Rain-kl/Wavelet/internal/model"
analyticsmodel "github.com/Rain-kl/Wavelet/internal/model/analytics"
)
// ErrMigrating 表示日志数据库正在迁移,当前禁止写入。
var ErrMigrating = errors.New("log database is migrating, writes are disabled")
// AccessLogStore 节点访问日志(of_node_access_logs)。
type AccessLogStore interface {
// InsertBatch 为写入入口:冻结检查 + 经 hook 入队(异步),不直接落库。
InsertBatch(ctx context.Context, records []*model.OpenFlareAccessLog) error
// BatchInsertNodeAccessLogs 为 batchwriter flush 目标:直接批量写入当前存储。
BatchInsertNodeAccessLogs(ctx context.Context, rows []analyticsmodel.NodeAccessLog) error
List(ctx context.Context, query model.OpenFlareAccessLogQuery) ([]*model.OpenFlareAccessLog, error)
Count(ctx context.Context, query model.OpenFlareAccessLogQuery) (int64, int64, int64, error)
RegionCounts(ctx context.Context, nodeID string, since time.Time, limit int) ([]*model.OpenFlareAccessLogRegionCount, error)
BucketAggregates(ctx context.Context, filter model.OpenFlareAccessLogQuery, bucketSeconds int64) ([]analyticsmodel.NodeAccessLogBucketAggregate, error)
CountBuckets(ctx context.Context, filter model.OpenFlareAccessLogQuery, bucketSeconds int64) (int64, error)
BucketDimensions(ctx context.Context, filter model.OpenFlareAccessLogQuery, column string, bucketSeconds int64) ([]analyticsmodel.NodeAccessLogBucketDimension, error)
IPAggregates(ctx context.Context, filter model.OpenFlareAccessLogQuery, exactRemoteAddr bool) ([]analyticsmodel.NodeAccessLogIPAggregate, error)
IPSummaries(ctx context.Context, filter model.OpenFlareAccessLogQuery, recentSince time.Time) ([]analyticsmodel.NodeAccessLogIPSummary, error)
CountIPSummaries(ctx context.Context, filter model.OpenFlareAccessLogQuery) (int64, error)
WAFIPAggregates(ctx context.Context, filter model.OpenFlareAccessLogQuery) ([]analyticsmodel.NodeAccessLogWAFIPAggregate, error)
IPTrend(ctx context.Context, filter model.OpenFlareAccessLogQuery, bucketSeconds int64) ([]analyticsmodel.NodeAccessLogIPTrend, error)
TrafficSummary(ctx context.Context, filter model.OpenFlareAccessLogQuery) (model.OpenFlareAccessLogTrafficSummary, error)
ValueCounts(ctx context.Context, filter model.OpenFlareAccessLogQuery, column string, limit int) ([]model.OpenFlareAccessLogValueCount, error)
NodeAggregates(ctx context.Context, filter model.OpenFlareAccessLogQuery) ([]model.OpenFlareAccessLogNodeAggregate, error)
DeleteAll(ctx context.Context) (int64, error)
DeleteBefore(ctx context.Context, cutoff time.Time) (int64, error)
DeleteByNodeBefore(ctx context.Context, nodeID string, before time.Time) (int64, error)
// ListForMigration 按 id 升序分页读取(迁移复制用)。
ListForMigration(ctx context.Context, afterID uint64, limit int) ([]analyticsmodel.NodeAccessLog, error)
// MigrationRange 返回源表 logged_at 的最小/最大值(空表返回零值),迁移预建分区用。
MigrationRange(ctx context.Context) (from, to time.Time, err error)
// EnsurePartitions 幂等预建 PG 分区(按月),覆盖 [from, to] 月份;CH/SQLite 为 no-op。
// 目标为 PG 的迁移在复制前调用,避免历史数据写入报 "no partition of relation found"。
EnsurePartitions(ctx context.Context, from, to time.Time) error
}
// ObservabilityStore 可观测 4 表(metric snapshots / edge health / frps / frpc)。
type ObservabilityStore interface {
InsertMetricSnapshot(ctx context.Context, record *model.OpenFlareMetricSnapshot) error
ListMetricSnapshots(ctx context.Context, nodeID string, since time.Time, limit int) ([]*model.OpenFlareMetricSnapshot, error)
DeleteAllMetricSnapshots(ctx context.Context) (int64, error)
DeleteMetricSnapshotsBefore(ctx context.Context, cutoff time.Time) (int64, error)
BatchInsertNodeMetricSnapshots(ctx context.Context, rows []analyticsmodel.NodeMetricSnapshot) error
// ListTrafficHourly 返回小时级流量汇总(按 node/hour 聚合,unique_visitor_count 恒 0)。
// CH 后端读 of_access_log_hourly rollup;PG/SQLite 从 of_node_access_logs 实时聚合。
ListTrafficHourly(ctx context.Context, nodeID string, since time.Time) ([]analyticsmodel.NodeTrafficHourly, error)
// ListAccessLogHourly 返回按 node/hour/host 的小时级访问日志汇总。
ListAccessLogHourly(ctx context.Context, nodeID string, since time.Time) ([]analyticsmodel.AccessLogHourly, error)
// ListMetricHourly 返回小时级指标聚合(avg cpu/memory + 计数器增量,reported_nodes 去重节点数)。
ListMetricHourly(ctx context.Context, nodeID string, since time.Time) ([]analyticsmodel.NodeMetricHourly, error)
InsertEdgeHealth(ctx context.Context, record *model.OpenFlareEdgeHealth) error
ListEdgeHealth(ctx context.Context, nodeID string, since time.Time, limit int) ([]*model.OpenFlareEdgeHealth, error)
DeleteAllEdgeHealth(ctx context.Context) (int64, error)
DeleteEdgeHealthBefore(ctx context.Context, cutoff time.Time) (int64, error)
BatchInsertNodeEdgeHealth(ctx context.Context, rows []analyticsmodel.NodeEdgeHealth) error
InsertNodeObservationFrps(ctx context.Context, record *model.OpenFlareNodeObservationFrps) error
ListNodeObservationFrps(ctx context.Context, nodeID string, since time.Time, limit int) ([]*model.OpenFlareNodeObservationFrps, error)
DeleteAllNodeObservationFrps(ctx context.Context) (int64, error)
DeleteNodeObservationFrpsBefore(ctx context.Context, cutoff time.Time) (int64, error)
BatchInsertNodeObsFrps(ctx context.Context, rows []analyticsmodel.NodeObsFrps) error
InsertNodeObservationFrpc(ctx context.Context, record *model.OpenFlareNodeObservationFrpc) error
ListNodeObservationFrpc(ctx context.Context, nodeID string, since time.Time, limit int) ([]*model.OpenFlareNodeObservationFrpc, error)
DeleteAllNodeObservationFrpc(ctx context.Context) (int64, error)
DeleteNodeObservationFrpcBefore(ctx context.Context, cutoff time.Time) (int64, error)
BatchInsertNodeObsFrpc(ctx context.Context, rows []analyticsmodel.NodeObsFrpc) error
// 迁移复制用:按 id 升序分页读取。
ListMetricSnapshotsForMigration(ctx context.Context, afterID uint64, limit int) ([]analyticsmodel.NodeMetricSnapshot, error)
ListEdgeHealthForMigration(ctx context.Context, afterID uint64, limit int) ([]analyticsmodel.NodeEdgeHealth, error)
ListNodeObsFrpsForMigration(ctx context.Context, afterID uint64, limit int) ([]analyticsmodel.NodeObsFrps, error)
ListNodeObsFrpcForMigration(ctx context.Context, afterID uint64, limit int) ([]analyticsmodel.NodeObsFrpc, error)
}
// UserAccessLogStore 用户访问日志(w_user_access_logs)。
type UserAccessLogStore interface {
BatchInsert(ctx context.Context, logs []analyticsmodel.UserAccessLog) error
// DeleteAll 清空全部用户访问日志(迁移「覆盖目标库已有日志」幂等前提用)。
DeleteAll(ctx context.Context) (int64, error)
Count(ctx context.Context, filter analyticsmodel.AccessLogFilter) (uint64, error)
List(ctx context.Context, filter analyticsmodel.AccessLogFilter, page, pageSize int) ([]analyticsmodel.UserAccessLog, uint64, error)
GetDailyTrend(ctx context.Context, days int) ([]analyticsmodel.DailyTrend, error)
GetBrowserDistribution(ctx context.Context, startTime time.Time) ([]analyticsmodel.BrowserShare, error)
GetTopActiveUsers(ctx context.Context, startTime time.Time, limit int) ([]analyticsmodel.TopUser, error)
// ListForMigration 按 id 升序分页读取(迁移复制用)。
ListForMigration(ctx context.Context, afterID uint64, limit int) ([]analyticsmodel.UserAccessLog, error)
// MigrationRange 返回源表 created_at 的最小/最大值(空表返回零值),迁移预建分区用。
MigrationRange(ctx context.Context) (from, to time.Time, err error)
}
// StatusStore 日志库状态(供管理端状态端点)。
type StatusStore interface {
ActiveDatabase(ctx context.Context) (string, error)
ClickHouseOperationalStats(ctx context.Context) (*analyticsmodel.ClickHouseOperationalStats, error) // 仅 CH 激活时非 nil
}
// Store 聚合当前生效日志库的全部域存储。
type Store struct {
AccessLogs AccessLogStore
Observability ObservabilityStore
UserAccessLogs UserAccessLogStore
Status StatusStore
}
@@ -0,0 +1,183 @@
// Copyright 2026 Arctel.net
// SPDX-License-Identifier: Apache-2.0
package logstore
import (
"context"
"fmt"
"os"
"regexp"
"strings"
"testing"
"time"
"gorm.io/driver/postgres"
"gorm.io/gorm"
"gorm.io/gorm/logger"
analyticsmodel "github.com/Rain-kl/Wavelet/internal/model/analytics"
)
// TestEnsurePartitionsPostgresInsertAcrossMonths 需要 TEST_POSTGRES_DSN(未设置时跳过):
// 验证 EnsurePartitions 预建任意月份范围分区后,跨月历史数据可写入 PG 分区表
// (对应迁移任务从 CH/SQLite 复制历史日志到 PG 时先预建分区的场景)。
func TestEnsurePartitionsPostgresInsertAcrossMonths(t *testing.T) {
dsn := strings.TrimSpace(os.Getenv("TEST_POSTGRES_DSN"))
if dsn == "" {
t.Skip("TEST_POSTGRES_DSN is not set")
}
gdb, err := gorm.Open(postgres.Open(dsn), &gorm.Config{
DisableForeignKeyConstraintWhenMigrating: true,
Logger: logger.Default.LogMode(logger.Silent),
})
if err != nil {
t.Fatalf("open postgres: %v", err)
}
sqlDB, err := gdb.DB()
if err != nil {
t.Fatalf("sql db: %v", err)
}
sqlDB.SetMaxOpenConns(1)
schema := fmt.Sprintf("logstore_partition_%d", time.Now().UnixNano())
if !regexp.MustCompile(`^[a-z0-9_]+$`).MatchString(schema) {
t.Fatalf("invalid schema: %s", schema)
}
if err := gdb.Exec(`CREATE SCHEMA "` + schema + `"`).Error; err != nil {
t.Fatalf("create schema: %v", err)
}
if err := gdb.Exec(`SET search_path TO "` + schema + `"`).Error; err != nil {
t.Fatalf("set search_path: %v", err)
}
t.Cleanup(func() {
_ = gdb.Exec("SET search_path TO public").Error
_ = gdb.Exec(`DROP SCHEMA IF EXISTS "` + schema + `" CASCADE`).Error
_ = sqlDB.Close()
})
// 与 goose/postgres/202608080001_create_log_tables.sql 保持一致的分区父表 DDL。
for _, ddl := range []string{postgresNodeAccessLogsDDL, postgresUserAccessLogsDDL} {
if err := gdb.Exec(ddl).Error; err != nil {
t.Fatalf("create partitioned table: %v", err)
}
}
ResetForTest()
SetConfigReader(func(_ context.Context, _ string) (string, error) { return "", nil })
defer ResetForTest()
ctx := context.Background()
store := newGormStore(gdb)
ua := newUserAccessLogGormStore(gdb)
// 源范围跨 3 个月:2026-01-10 ~ 2026-03-20;to+1 月兜底生成 202601..202604 分区。
from := time.Date(2026, 1, 10, 8, 0, 0, 0, time.UTC)
max := time.Date(2026, 3, 20, 9, 30, 0, 0, time.UTC)
if err := store.EnsurePartitions(ctx, from, max.AddDate(0, 1, 0)); err != nil {
t.Fatalf("EnsurePartitions: %v", err)
}
// 幂等:重复调用不报错(CREATE TABLE IF NOT EXISTS ... PARTITION OF)。
if err := store.EnsurePartitions(ctx, from, max.AddDate(0, 1, 0)); err != nil {
t.Fatalf("EnsurePartitions idempotent: %v", err)
}
var partitionCount int64
if err := gdb.Raw(
"SELECT count(*) FROM pg_inherits WHERE inhrelid = to_regclass('of_node_access_logs')",
).Scan(&partitionCount).Error; err != nil {
t.Fatalf("count partitions: %v", err)
}
if partitionCount != 4 {
t.Fatalf("of_node_access_logs partitions = %d, want 4", partitionCount)
}
// 跨月插入:1/2/3 月各 2 条节点访问日志 + 2 条用户访问日志,均应命中已有分区。
nodeRows := []analyticsmodel.NodeAccessLog{
{ID: 1, NodeID: "n1", LoggedAt: time.Date(2026, 1, 15, 0, 0, 0, 0, time.UTC), RemoteAddr: "1.1.1.1"},
{ID: 2, NodeID: "n1", LoggedAt: time.Date(2026, 1, 20, 0, 0, 0, 0, time.UTC), RemoteAddr: "1.1.1.2"},
{ID: 3, NodeID: "n2", LoggedAt: time.Date(2026, 2, 10, 0, 0, 0, 0, time.UTC), RemoteAddr: "2.2.2.2"},
{ID: 4, NodeID: "n2", LoggedAt: time.Date(2026, 2, 12, 0, 0, 0, 0, time.UTC), RemoteAddr: "2.2.2.3"},
{ID: 5, NodeID: "n1", LoggedAt: time.Date(2026, 3, 5, 0, 0, 0, 0, time.UTC), RemoteAddr: "3.3.3.3"},
{ID: 6, NodeID: "n1", LoggedAt: time.Date(2026, 3, 18, 0, 0, 0, 0, time.UTC), RemoteAddr: "3.3.3.4"},
}
if err := store.BatchInsertNodeAccessLogs(ctx, nodeRows); err != nil {
t.Fatalf("insert node access logs across months: %v", err)
}
userRows := []analyticsmodel.UserAccessLog{
{ID: 1, UserID: 101, Path: "/a", CreatedAt: time.Date(2026, 1, 16, 0, 0, 0, 0, time.UTC)},
{ID: 2, UserID: 102, Path: "/b", CreatedAt: time.Date(2026, 3, 17, 0, 0, 0, 0, time.UTC)},
}
if err := ua.BatchInsert(ctx, userRows); err != nil {
t.Fatalf("insert user access logs across months: %v", err)
}
var nodeCount, userCount int64
if err := gdb.Model(&analyticsmodel.NodeAccessLog{}).Count(&nodeCount).Error; err != nil {
t.Fatalf("count node access logs: %v", err)
}
if err := gdb.Model(&analyticsmodel.UserAccessLog{}).Count(&userCount).Error; err != nil {
t.Fatalf("count user access logs: %v", err)
}
if nodeCount != 6 {
t.Fatalf("node access log count = %d, want 6", nodeCount)
}
if userCount != 2 {
t.Fatalf("user access log count = %d, want 2", userCount)
}
// MigrationRange 返回跨月范围(覆盖两表)。
gotFrom, gotTo, err := store.MigrationRange(ctx)
if err != nil {
t.Fatalf("node MigrationRange: %v", err)
}
if !gotFrom.Equal(time.Date(2026, 1, 15, 0, 0, 0, 0, time.UTC)) || !gotTo.Equal(time.Date(2026, 3, 18, 0, 0, 0, 0, time.UTC)) {
t.Fatalf("node MigrationRange = %s ~ %s, want 2026-01-15 ~ 2026-03-18", gotFrom, gotTo)
}
uaFrom, uaTo, err := ua.MigrationRange(ctx)
if err != nil {
t.Fatalf("user MigrationRange: %v", err)
}
if !uaFrom.Equal(time.Date(2026, 1, 16, 0, 0, 0, 0, time.UTC)) || !uaTo.Equal(time.Date(2026, 3, 17, 0, 0, 0, 0, time.UTC)) {
t.Fatalf("user MigrationRange = %s ~ %s", uaFrom, uaTo)
}
}
// postgresNodeAccessLogsDDL 与 goose/postgres/202608080001_create_log_tables.sql 对齐。
const postgresNodeAccessLogsDDL = `
CREATE TABLE IF NOT EXISTS of_node_access_logs (
id BIGINT NOT NULL,
node_id VARCHAR(64) NOT NULL DEFAULT '',
logged_at TIMESTAMPTZ NOT NULL,
remote_addr VARCHAR(128) NOT NULL DEFAULT '',
region VARCHAR(128) NOT NULL DEFAULT '',
host VARCHAR(255) NOT NULL DEFAULT '',
path VARCHAR(2048) NOT NULL DEFAULT '',
user_agent TEXT NOT NULL DEFAULT '',
cache_status VARCHAR(64) NOT NULL DEFAULT '',
status_code INTEGER NOT NULL DEFAULT 0,
bytes_sent BIGINT NOT NULL DEFAULT 0,
request_length BIGINT NOT NULL DEFAULT 0,
request_time_ms INTEGER NOT NULL DEFAULT 0,
created_at TIMESTAMPTZ NOT NULL DEFAULT CURRENT_TIMESTAMP,
PRIMARY KEY (id, logged_at)
) PARTITION BY RANGE (logged_at)`
// postgresUserAccessLogsDDL 与 goose/postgres/202608080001_create_log_tables.sql 对齐。
const postgresUserAccessLogsDDL = `
CREATE TABLE IF NOT EXISTS w_user_access_logs (
id BIGINT NOT NULL,
user_id BIGINT NOT NULL DEFAULT 0,
path VARCHAR(2048) NOT NULL DEFAULT '',
method VARCHAR(16) NOT NULL DEFAULT '',
ip VARCHAR(128) NOT NULL DEFAULT '',
user_agent TEXT NOT NULL DEFAULT '',
headers TEXT NOT NULL DEFAULT '',
status INTEGER NOT NULL DEFAULT 0,
latency BIGINT NOT NULL DEFAULT 0,
created_at TIMESTAMPTZ NOT NULL DEFAULT CURRENT_TIMESTAMP,
PRIMARY KEY (id, created_at)
) PARTITION BY RANGE (created_at)`
File diff suppressed because it is too large Load Diff
File diff suppressed because it is too large Load Diff
+170
View File
@@ -0,0 +1,170 @@
// Copyright 2026 Arctel.net
// SPDX-License-Identifier: Apache-2.0
package logstore
import (
"context"
"errors"
"fmt"
"sync"
"github.com/Rain-kl/Wavelet/internal/infra/config"
db "github.com/Rain-kl/Wavelet/internal/infra/persistence"
"github.com/Rain-kl/Wavelet/internal/model"
)
// logDatabaseKey / logMigrationKey 对应 model.ConfigKeyLogDatabase / ConfigKeyLogDBMigration。
const (
logDatabaseKey = model.ConfigKeyLogDatabase
logMigrationKey = model.ConfigKeyLogDBMigration
)
// 日志库名常量(与 model 配置值一致,集中避免散落字符串字面量)。
const (
dbNamePostgres = "postgres"
dbNameSQLite = "sqlite"
dbNameClickHouse = "clickhouse"
)
// errConfigReaderNotWired 表示 config reader 尚未注入(首启/测试场景按 seed 规则兜底)。
var errConfigReaderNotWired = errors.New("logstore: config reader not wired")
// ConfigReader 读取系统配置字符串值,由 bootstrap 注入(避免 logstore ↔ repository 循环依赖)。
type ConfigReader func(ctx context.Context, key string) (string, error)
var (
configReader ConfigReader
storeMu sync.RWMutex
active *Store
activeDB string
)
// SetConfigReader 注入系统配置读取函数(bootstrap 调用,测试可注入内存实现)。
func SetConfigReader(fn ConfigReader) { configReader = fn }
func getConfig(ctx context.Context, key string) (string, error) {
if configReader == nil {
return "", errConfigReaderNotWired
}
return configReader(ctx, key)
}
// Active 返回当前生效的日志库 Store。按 log_database 系统配置惰性解析并缓存,
// 配置更新(含迁移任务翻转)后自动重建。
func Active(ctx context.Context) (*Store, error) {
current, err := resolveDatabase(ctx)
if err != nil {
return nil, err
}
storeMu.RLock()
if active != nil && activeDB == current {
s := active
storeMu.RUnlock()
return s, nil
}
storeMu.RUnlock()
storeMu.Lock()
defer storeMu.Unlock()
if active != nil && activeDB == current {
return active, nil
}
s, err := buildStore(ctx, current, false)
if err != nil {
return nil, err
}
active = s
activeDB = current
return s, nil
}
// Build 直接按目标构造 store(不经 Active 缓存)。
func Build(ctx context.Context, database string) (*Store, error) {
return buildStore(ctx, database, false)
}
// BuildForMigration 构造迁移目标 store:与 Build 相同但不做冻结检查
// (迁移期间 log_db_migration=migrating 已冻结源库写入,目标库的清空/复制写入必须放行)。
func BuildForMigration(ctx context.Context, database string) (*Store, error) {
return buildStore(ctx, database, true)
}
// buildStore 按目标构造实现。skipFreeze 为 true 时该 store 跳过冻结检查
// (仅迁移任务的目标 store 使用)。gorm 分支 UserAccessLogs 用独立包装类型
// (gormLogStore 已占用 List/Count 方法名,无法再实现 UserAccessLogStore)。
func buildStore(ctx context.Context, database string, skipFreeze bool) (*Store, error) {
switch database {
case dbNameClickHouse:
ch := newClickHouseStore()
ch.skipFreeze = skipFreeze
return &Store{
AccessLogs: ch,
Observability: ch,
UserAccessLogs: newClickHouseUserAccessLogStore(),
Status: ch,
}, nil
case dbNamePostgres, dbNameSQLite:
gdb := db.DB(ctx)
g := newGormStore(gdb)
g.skipFreeze = skipFreeze
return &Store{
AccessLogs: g,
Observability: g,
UserAccessLogs: newUserAccessLogGormStore(gdb),
Status: g,
}, nil
default:
return nil, fmt.Errorf("unsupported log database: %s", database)
}
}
// Migrating 返回日志库是否处于迁移冻结状态。
func Migrating(ctx context.Context) bool {
v, err := getConfig(ctx, logMigrationKey)
if err != nil {
return false
}
return v == "migrating"
}
// Init 在 bootstrap 阶段预热一次激活 store(幂等,失败不致命——首次使用时再解析)。
func Init(ctx context.Context) {
_, _ = Active(ctx)
}
// ResetForTest 清空缓存的激活 store 与 config reader,便于测试注入。
func ResetForTest() {
storeMu.Lock()
active = nil
activeDB = ""
storeMu.Unlock()
configReader = nil
}
// ActiveDatabase 返回当前日志主库名(postgres|sqlite|clickhouse)。
func ActiveDatabase(ctx context.Context) (string, error) {
return resolveDatabase(ctx)
}
// resolveDatabase 读取 log_database:值缺失或 reader 未装配(首启)时按启动规则 seed;
// 已装配 reader 的真实读取错误直接透出,避免把读失败当首次启动。
func resolveDatabase(ctx context.Context) (string, error) {
v, err := getConfig(ctx, logDatabaseKey)
if err != nil && !errors.Is(err, errConfigReaderNotWired) {
return "", err
}
if v != "" {
return v, nil
}
// 首次启动 seed:CH 启用 → clickhouse;否则随主库。
defaultDB := dbNameSQLite
if config.Config.Database.Enabled {
defaultDB = dbNamePostgres
}
if config.Config.ClickHouse.Enabled {
defaultDB = dbNameClickHouse
}
return defaultDB, nil
}
@@ -0,0 +1,109 @@
// Copyright 2026 Arctel.net
// SPDX-License-Identifier: Apache-2.0
package logstore
import (
"context"
"errors"
"testing"
"time"
analyticsmodel "github.com/Rain-kl/Wavelet/internal/model/analytics"
)
func TestMigratingReadsConfig(t *testing.T) {
ResetForTest()
SetConfigReader(func(_ context.Context, key string) (string, error) {
if key == logMigrationKey {
return "migrating", nil
}
return "", nil
})
if !Migrating(context.Background()) {
t.Fatal("Migrating() = false, want true when key=migrating")
}
SetConfigReader(func(_ context.Context, key string) (string, error) {
return "", nil
})
if Migrating(context.Background()) {
t.Fatal("Migrating() = true, want false when key empty")
}
}
func TestResolveDatabaseDefaults(t *testing.T) {
ResetForTest()
// 配置缺失(reader 返回空值)时按主库规则 seed(config.Config 默认值由既有测试基建决定)。
SetConfigReader(func(_ context.Context, key string) (string, error) {
return "", nil
})
got, err := resolveDatabase(context.Background())
if err != nil {
t.Fatalf("resolveDatabase: %v", err)
}
if got != "postgres" && got != "sqlite" && got != "clickhouse" {
t.Fatalf("unexpected default log database: %s", got)
}
}
func TestResolveDatabaseSurfacesReadError(t *testing.T) {
ResetForTest()
wantErr := errors.New("boom")
SetConfigReader(func(_ context.Context, key string) (string, error) {
return "", wantErr
})
if _, err := resolveDatabase(context.Background()); !errors.Is(err, wantErr) {
t.Fatalf("resolveDatabase error = %v, want %v", err, wantErr)
}
}
func TestActiveBuildsStore(t *testing.T) {
ResetForTest()
SetConfigReader(func(_ context.Context, key string) (string, error) {
if key == logDatabaseKey {
return "sqlite", nil
}
return "", nil
})
store, err := Active(context.Background())
if err != nil {
t.Fatalf("Active: %v", err)
}
if store == nil {
t.Fatal("Active() returned nil store")
}
if store.AccessLogs == nil || store.Observability == nil || store.UserAccessLogs == nil || store.Status == nil {
t.Fatalf("Active() store fields not fully wired: %+v", store)
}
// 再次调用应命中缓存。
again, err := Active(context.Background())
if err != nil {
t.Fatalf("Active (cached): %v", err)
}
if again != store {
t.Fatal("Active() did not return cached store")
}
}
// TestClickHouseUserAccessLogBatchInsertFreeze 覆盖 CH 用户访问日志 flush 的冻结检查:
// 冻结期非空批次返回 ErrMigrating(在触碰 CH 连接之前),空批次直接成功。
func TestClickHouseUserAccessLogBatchInsertFreeze(t *testing.T) {
ResetForTest()
SetConfigReader(func(_ context.Context, key string) (string, error) {
if key == logMigrationKey {
return "migrating", nil
}
return "", nil
})
defer ResetForTest()
s := newClickHouseUserAccessLogStore()
ctx := context.Background()
now := time.Now()
if err := s.BatchInsert(ctx, []analyticsmodel.UserAccessLog{{UserID: 1, CreatedAt: now}}); !errors.Is(err, ErrMigrating) {
t.Fatalf("BatchInsert during migration: want ErrMigrating, got %v", err)
}
if err := s.BatchInsert(ctx, nil); err != nil {
t.Fatalf("BatchInsert empty batch: %v", err)
}
}