diff --git a/backend/controllers/alerts.go b/backend/controllers/alerts.go index dca4959..260d0fd 100644 --- a/backend/controllers/alerts.go +++ b/backend/controllers/alerts.go @@ -182,10 +182,18 @@ func (ctrl *AlertsController) CheckNow(c *gin.Context) { // Deliveries 发送历史(最近 50) func (ctrl *AlertsController) Deliveries(c *gin.Context) { + // 保留期由系统设置控制（默认 7 天），库里已是裁剪后的数据； + // 这里只是限制单次响应大小，避免历史多了以后一次吐出上万条。 + limit := 500 + if raw := c.Query("limit"); raw != "" { + if v, err := strconv.Atoi(raw); err == nil && v > 0 && v <= 2000 { + limit = v + } + } db := repository.GetDB() db.RLock() - out := make([]map[string]interface{}, 0, len(db.AlertDeliveries)) - for i := len(db.AlertDeliveries) - 1; i >= 0 && len(out) < 50; i-- { + out := make([]map[string]interface{}, 0, limit) + for i := len(db.AlertDeliveries) - 1; i >= 0 && len(out) < limit; i-- { d := db.AlertDeliveries[i] out = append(out, map[string]interface{}{ "id": d.ID, "ruleId": d.RuleID, "ruleName": d.RuleName, diff --git a/backend/controllers/system.go b/backend/controllers/system.go index d4704cf..6d83b89 100644 --- a/backend/controllers/system.go +++ b/backend/controllers/system.go @@ -84,6 +84,25 @@ func (ctrl *SystemController) Backups(c *gin.Context) { utils.Success(c, gin.H{"list": services.ListBackups()}) } +// GetRetention 读取平台数据保留期（告警发送历史 / 审计日志）。 +func (ctrl *SystemController) GetRetention(c *gin.Context) { + utils.Success(c, services.GetRetentionSettings()) +} + +// SaveRetention 保存平台数据保留期。 +// 校验范围而不是直接接受：保留 0 天等于"什么都不留"，通常是误操作。 +func (ctrl *SystemController) SaveRetention(c *gin.Context) { + var req services.RetentionSettings + if !utils.BindAndValidate(c, &req) { + return + } + if err := services.SaveRetentionSettings(req.AlertDeliveryDays, req.AuditLogDays); err != nil { + utils.BadRequest(c, err.Error()) + return + } + utils.Success(c, services.GetRetentionSettings()) +} + // Health 平台自检：一键诊断用。 // 聚合「能不能用」的判断，避免排障时在多个页面之间跳。 func (ctrl *SystemController) Health(c *gin.Context) { diff --git a/backend/routers/router.go b/backend/routers/router.go index 2585696..452be82 100644 --- a/backend/routers/router.go +++ b/backend/routers/router.go @@ -105,6 +105,10 @@ func SetupRouter() *gin.Engine { logs.GET("/health", logsCtrl.Health) } + // 平台数据保留期（告警发送历史 / 审计日志） + authorized.GET("/settings/retention", systemCtrl.GetRetention) + authorized.PUT("/settings/retention", middleware.AdminAuth(), systemCtrl.SaveRetention) + // 镜像源(国内加速/海外/私有仓库) authorized.GET("/settings/image-registry", registryCtrl.Get) authorized.PUT("/settings/image-registry", middleware.AdminAuth(), registryCtrl.Save) diff --git a/backend/services/alert.go b/backend/services/alert.go index 72e9966..498ac83 100644 --- a/backend/services/alert.go +++ b/backend/services/alert.go @@ -227,6 +227,37 @@ func recordCooldownThrottled(ruleID uint) bool { return true } +// pruneDeliveriesLocked 按保留期 + 条数上限裁剪投递历史（调用方必须已持锁）。 +func pruneDeliveriesLocked(db *repository.JsonDB) { + cut := 0 + // 注意：调用方已持写锁，必须用无锁读取版本，否则 RWMutex 会死锁 + deadline := time.Now().Add(-alertDeliveryRetentionLocked(db)) + for cut < len(db.AlertDeliveries) && db.AlertDeliveries[cut].CreatedAt.Before(deadline) { + cut++ + } + if remain := len(db.AlertDeliveries) - cut; remain > maxAlertDeliveries { + cut += remain - maxAlertDeliveries + } + if cut > 0 { + db.AlertDeliveries = append([]models.AlertDelivery{}, db.AlertDeliveries[cut:]...) + } +} + +// PruneAlertDeliveries 按当前保留期裁剪一次（供保存设置后立即生效）。 +func PruneAlertDeliveries() { + db := repository.GetDB() + if db == nil { + return + } + db.Lock() + defer db.Unlock() + before := len(db.AlertDeliveries) + pruneDeliveriesLocked(db) + if len(db.AlertDeliveries) != before { + db.RequestSave() + } +} + func recordDelivery(ruleID uint, ruleName string, count int, status, reason string) { db := repository.GetDB() newID := db.NextID("alertDeliveries") @@ -235,10 +266,9 @@ func recordDelivery(ruleID uint, ruleName string, count int, status, reason stri ID: newID, RuleID: ruleID, RuleName: ruleName, Count: count, Status: status, Reason: reason, }) - // 只保留最近 50 条 - if len(db.AlertDeliveries) > 50 { - db.AlertDeliveries = db.AlertDeliveries[len(db.AlertDeliveries)-50:] - } + // 保留策略：按 告警发送历史保留天数（默认 7 天，系统设置可改）+ 条数硬上限兜底。 + // 原来写死 只留最近 50 条，真实发送记录一小时内就被冷却记录挤光了。 + pruneDeliveriesLocked(db) db.RequestSave() db.Unlock() } diff --git a/backend/services/audit.go b/backend/services/audit.go index a7d4ef8..e05de09 100644 --- a/backend/services/audit.go +++ b/backend/services/audit.go @@ -15,8 +15,7 @@ const ( // maxAuditLogs 审计日志条数上限。JSON DB 是全量读进内存、全量写盘， // 不设上限的话审计表会把内存和落盘时间一起拖垮。 maxAuditLogs = 20000 - // auditRetention 审计日志保留天数（合规常见要求 90 天） - auditRetention = 90 * 24 * time.Hour + // 保留期改由系统设置配置（见 services/retention.go） ) // RecordAudit 写入一条审计日志。 @@ -60,7 +59,9 @@ func RecordAudit(entry models.AuditLog) { func pruneAuditLocked(db *repository.JsonDB) { logs := db.AuditLogs cut := 0 - deadline := time.Now().Add(-auditRetention) + // 保留期来自 系统设置->通用配置->日志保留时间，默认 90 天 + // 注意：调用方已持写锁，必须用无锁读取版本，否则 RWMutex 会死锁 + deadline := time.Now().Add(-auditLogRetentionLocked(db)) // 1) 先按保留期裁 for cut < len(logs) && logs[cut].CreatedAt.Before(deadline) { @@ -75,6 +76,21 @@ func pruneAuditLocked(db *repository.JsonDB) { } } +// PruneAuditLogs 按当前保留期裁剪一次审计日志（供保存设置后立即生效）。 +func PruneAuditLogs() { + db := repository.GetDB() + if db == nil { + return + } + db.Lock() + defer db.Unlock() + before := len(db.AuditLogs) + pruneAuditLocked(db) + if len(db.AuditLogs) != before { + db.RequestSave() + } +} + // ListAuditLogs 分页查询审计日志（按时间倒序）。 func ListAuditLogs(q models.AuditQuery) ([]models.AuditLog, int64) { db := repository.GetDB() diff --git a/backend/services/cluster_workloads_test.go b/backend/services/cluster_workloads_test.go index 91e9371..b50d438 100644 --- a/backend/services/cluster_workloads_test.go +++ b/backend/services/cluster_workloads_test.go @@ -3,9 +3,9 @@ package services import ( "context" "fmt" + "lokistack/models" "net/http" "net/http/httptest" - "lokistack/models" "testing" ) diff --git a/backend/services/collector_templates.go b/backend/services/collector_templates.go index e2b5ee1..79201fa 100644 --- a/backend/services/collector_templates.go +++ b/backend/services/collector_templates.go @@ -15,6 +15,7 @@ import ( // indent 必须与该模板里 `external_labels:` 的层级匹配： // - 主机模板（4 空格）→ 子项 6 空格 // - DaemonSet 模板（6 空格）→ 子项 8 空格 +// // 早先写死 8 空格，主机模式一旦配了标签就会生成非法 YAML，Promtail 直接装不上。 func promtailExternalLabels(cluster string, labelsJSON string, indent string) string { lines := []string{fmt.Sprintf("%scluster: %s", indent, yamlScalar(cluster))} diff --git a/backend/services/compactor_workers_test.go b/backend/services/compactor_workers_test.go index b7f3734..b3e28c0 100644 --- a/backend/services/compactor_workers_test.go +++ b/backend/services/compactor_workers_test.go @@ -11,13 +11,13 @@ import ( // 统一一个值要么被限流（COS 503 SlowDown），要么白等（S3 能扛更多）。 func TestCompactorDeleteWorkersPerStorageType(t *testing.T) { cases := map[string]int{ - "cos": 16, // 实测 150 触发 503 SlowDown - "obs": 32, // 华为 OBS 保守值 - "minio": 32, // 自建，取决于磁盘 - "azure": 32, // 存储账户限流较严 - "oss": 64, // 阿里云 OSS QPS 较高 + "cos": 16, // 实测 150 触发 503 SlowDown + "obs": 32, // 华为 OBS 保守值 + "minio": 32, // 自建，取决于磁盘 + "azure": 32, // 存储账户限流较严 + "oss": 64, // 阿里云 OSS QPS 较高 "gcs": 64, - "s3": 96, // AWS S3 限流宽松 + "s3": 96, // AWS S3 限流宽松 "": defaultCompactorDeleteWorkers, "unknown": defaultCompactorDeleteWorkers, } diff --git a/backend/services/k8s_client.go b/backend/services/k8s_client.go index 15a17d9..65a4ca7 100644 --- a/backend/services/k8s_client.go +++ b/backend/services/k8s_client.go @@ -7,10 +7,10 @@ import ( "regexp" "time" - "lokistack/models" "k8s.io/client-go/kubernetes" "k8s.io/client-go/rest" "k8s.io/client-go/tools/clientcmd" + "lokistack/models" ) // resolveKubeconfig 取集群最终的 kubeconfig 文本（优先内存字段，其次磁盘文件） diff --git a/backend/services/k8s_resources.go b/backend/services/k8s_resources.go index 908b3cf..b252106 100644 --- a/backend/services/k8s_resources.go +++ b/backend/services/k8s_resources.go @@ -301,7 +301,6 @@ func patchDeployment(ctx context.Context, cluster *models.Cluster, namespace, na return err } - // podDNSBroken 在指定 deployment 的容器里探测 hostname 是否可解析(wget /ready 连通性)。 func podDNSBroken(ctx context.Context, cluster *models.Cluster, namespace, deployment, container, host string) bool { pods, err := listDeploymentPods(ctx, cluster, namespace, deployment) @@ -355,7 +354,6 @@ func listDeploymentPods(ctx context.Context, cluster *models.Cluster, namespace, return names, nil } - // lokiNamespaceOf 从 endpoint 的 svc FQDN 提取 Loki 命名空间(log.ns.svc → ns)。 func lokiNamespaceOf(loki *models.LokiDeployment, endpoint string) string { if loki != nil && loki.Namespace != "" { diff --git a/backend/services/loki_config.go b/backend/services/loki_config.go index 5a67637..483dc37 100644 --- a/backend/services/loki_config.go +++ b/backend/services/loki_config.go @@ -123,12 +123,12 @@ type lokiConfigData struct { CompactorStoreType string // CompactorDeleteWorkers 过期 chunk 删除并发，按对象存储类型取默认值 CompactorDeleteWorkers int - RingInstanceAddr string // 单副本 127.0.0.1；多副本 0.0.0.0（须可被对端寻址） - RingKVStore string // 单副本 inmemory；多副本 memberlist - MemberlistBlock string // 多副本时非空，注入 memberlist 顶级配置 - RulerStorageBlock string // 对象存储时非空，注入 ruler.storage.s3/gcs/azure 嵌套配置 - MultiReplica bool - FilesystemStorage bool + RingInstanceAddr string // 单副本 127.0.0.1；多副本 0.0.0.0（须可被对端寻址） + RingKVStore string // 单副本 inmemory；多副本 memberlist + MemberlistBlock string // 多副本时非空，注入 memberlist 顶级配置 + RulerStorageBlock string // 对象存储时非空，注入 ruler.storage.s3/gcs/azure 嵌套配置 + MultiReplica bool + FilesystemStorage bool } func (d *LokiDeployer) generateLokiConfig() string { @@ -181,20 +181,20 @@ func (d *LokiDeployer) generateLokiConfig() string { } data := lokiConfigData{ - AuthEnabled: d.deployment.AuthEnabled, - ReplicationFactor: d.replicationFactor(), - RetentionPeriod: retentionHours, - StorageConfigBlock: storageBlock, - ObjectStore: objectStore, - RulerStoreType: rulerStore, + AuthEnabled: d.deployment.AuthEnabled, + ReplicationFactor: d.replicationFactor(), + RetentionPeriod: retentionHours, + StorageConfigBlock: storageBlock, + ObjectStore: objectStore, + RulerStoreType: rulerStore, CompactorStoreType: compactorStore, CompactorDeleteWorkers: compactorDeleteWorkers(d.deployment.StorageConfig), - RingInstanceAddr: ringAddr, - RingKVStore: ringStore, - MemberlistBlock: memberlistBlock, - RulerStorageBlock: d.buildRulerStorageBlock(), - MultiReplica: d.deployment.Replicas >= 2, - FilesystemStorage: objectStore == "filesystem", + RingInstanceAddr: ringAddr, + RingKVStore: ringStore, + MemberlistBlock: memberlistBlock, + RulerStorageBlock: d.buildRulerStorageBlock(), + MultiReplica: d.deployment.Replicas >= 2, + FilesystemStorage: objectStore == "filesystem", } tmpl, err := template.New("loki-config").Parse(lokiConfigTemplate) diff --git a/backend/services/retention.go b/backend/services/retention.go new file mode 100644 index 0000000..8ee8968 --- /dev/null +++ b/backend/services/retention.go @@ -0,0 +1,132 @@ +package services + +import ( + "errors" + "strconv" + "time" + + "lokistack/models" + "lokistack/repository" +) + +// 平台自身数据的保留期设置（不是 Loki 日志的保留期）。 +// +// 为什么需要可配：不同团队对"告警发送历史留多久 / 审计日志留多久"要求不同 — +// 合规场景要 90~180 天，小团队 7 天就够。写死在代码里只能改一次发一次版。 +// +// 注意区分：Loki 的日志保留期是**部署级**配置（组件部署 → 保留策略）， +// 这里管的是平台自己产生的小数据（投递历史、审计日志）。 +const ( + settingKeyAlertDeliveryDays = "alertDeliveryRetentionDays" + settingKeyAuditLogDays = "auditLogRetentionDays" + + defaultAlertDeliveryDays = 7 + defaultAuditLogDays = 90 + + // 硬上限：即使配置得很大，也不能让 JSON 文件型 DB 无界增长 + maxAlertDeliveries = 20000 +) + +// RetentionSettings 平台数据保留期（天） +type RetentionSettings struct { + AlertDeliveryDays int `json:"alertDeliveryDays"` + AuditLogDays int `json:"auditLogDays"` +} + +// GetRetentionSettings 读取保留期配置，未配置时返回默认值。 +func GetRetentionSettings() RetentionSettings { + db := repository.GetDB() + if db == nil { + return retentionDefaults() + } + db.RLock() + defer db.RUnlock() + return retentionFromDBLocked(db) +} + +func retentionDefaults() RetentionSettings { + return RetentionSettings{ + AlertDeliveryDays: defaultAlertDeliveryDays, + AuditLogDays: defaultAuditLogDays, + } +} + +// retentionFromDBLocked 从已持有的锁内读取配置，**不会再加锁**。 +// +// 为什么必须单独一个函数：Go 的 sync.RWMutex 不可重入。 +// 调用方已经 db.Lock() 时再调 GetRetentionSettings()（内部 RLock）会**永久死锁** —— +// 实测把整个平台写路径卡死（登录都要写审计，于是登录 28 秒返回空）。 +func retentionFromDBLocked(db *repository.JsonDB) RetentionSettings { + out := retentionDefaults() + for _, item := range db.Settings { + switch item.Key { + case settingKeyAlertDeliveryDays: + if v, err := strconv.Atoi(item.Value); err == nil && v > 0 { + out.AlertDeliveryDays = v + } + case settingKeyAuditLogDays: + if v, err := strconv.Atoi(item.Value); err == nil && v > 0 { + out.AuditLogDays = v + } + } + } + return out +} + +// SaveRetentionSettings 保存保留期配置（天数）。 +func SaveRetentionSettings(alertDays, auditDays int) error { + if alertDays < 1 || alertDays > 365 { + return errors.New("告警发送历史保留天数需在 1~365 之间") + } + if auditDays < 1 || auditDays > 3650 { + return errors.New("审计日志保留天数需在 1~3650 之间") + } + db := repository.GetDB() + if db == nil { + return errors.New("数据库未初始化") + } + db.Lock() + upsertSettingLocked(db, settingKeyAlertDeliveryDays, strconv.Itoa(alertDays)) + upsertSettingLocked(db, settingKeyAuditLogDays, strconv.Itoa(auditDays)) + db.RequestSave() + db.Unlock() + + // 立刻按新策略裁剪一次，避免"改了配置但旧数据还在" + PruneAlertDeliveries() + PruneAuditLogs() + return nil +} + +// upsertSettingLocked 写入/更新一条设置（调用方必须已持写锁）。 +func upsertSettingLocked(db *repository.JsonDB, key, value string) { + for i := range db.Settings { + if db.Settings[i].Key == key { + db.Settings[i].Value = value + return + } + } + // 不能用 db.NextID()：它内部会 d.mu.Lock()，而调用方已经持有写锁， + // sync.Mutex 不可重入 → 永久死锁（这个坑在审计日志里踩过一次）。 + // 这里在锁内自己算最大 ID + 1。 + nextID := uint(1) + for _, item := range db.Settings { + if item.ID >= nextID { + nextID = item.ID + 1 + } + } + db.Settings = append(db.Settings, models.SystemSetting{ + ID: nextID, + Key: key, + Value: value, + }) +} + +// alertDeliveryRetentionLocked 返回投递历史保留时长（调用方必须已持锁）。 +func alertDeliveryRetentionLocked(db *repository.JsonDB) time.Duration { + return time.Duration(retentionFromDBLocked(db).AlertDeliveryDays) * 24 * time.Hour +} + +// auditLogRetentionLocked 返回审计日志保留时长（调用方必须已持锁）。 +func auditLogRetentionLocked(db *repository.JsonDB) time.Duration { + return time.Duration(retentionFromDBLocked(db).AuditLogDays) * 24 * time.Hour +} diff --git a/backend/services/retention_test.go b/backend/services/retention_test.go new file mode 100644 index 0000000..2defcad --- /dev/null +++ b/backend/services/retention_test.go @@ -0,0 +1,51 @@ +package services + +import ( + "testing" + "time" + + "lokistack/repository" +) + +// 回归：持写锁时裁剪逻辑绝不能再去拿读锁。 +// +// 线上事故：一个 day-retention 配置读取被写成在 db.Lock() 内部调用 +// GetRetentionSettings()（内部 db.RLock()）。Go 的 sync.RWMutex 不可重入， +// 于是**第一次写投递历史/审计日志就永久死锁** —— 平台表现是"登录 28 秒返回空"， +// 因为登录也要写审计。go build / go vet / 单测（不触发写路径）全都查不出来。 +func TestPruneWhileLockedDoesNotDeadlock(t *testing.T) { + db := &repository.JsonDB{} + db.Lock() + defer db.Unlock() + + done := make(chan struct{}) + go func() { + defer close(done) + pruneDeliveriesLocked(db) + }() + + select { + case <-done: + case <-time.After(2 * time.Second): + t.Fatal("pruneDeliveriesLocked 在持有写锁时死锁（内部又去拿了读锁）") + } +} + +// 同理：审计日志裁剪。 +func TestPruneAuditWhileLockedDoesNotDeadlock(t *testing.T) { + db := &repository.JsonDB{} + db.Lock() + defer db.Unlock() + + done := make(chan struct{}) + go func() { + defer close(done) + pruneAuditLocked(db) + }() + + select { + case <-done: + case <-time.After(2 * time.Second): + t.Fatal("pruneAuditLocked 在持有写锁时死锁（内部又去拿了读锁）") + } +} diff --git a/backend/services/storage_check.go b/backend/services/storage_check.go index cd4b707..4158bc0 100644 --- a/backend/services/storage_check.go +++ b/backend/services/storage_check.go @@ -177,8 +177,8 @@ func stripBucketPrefix(endpoint, bucket string) string { // bucketPathFor 根据 endpoint 风格决定桶路径（用于建桶 PUT）。 // -// 1) 区域端点（推荐）: cos.ap-hongkong.myqcloud.com → 路径式 / -// 2) 虚拟主机式: mybucket-1250000000.cos.… → 路径为 / +// 1. 区域端点（推荐）: cos.ap-hongkong.myqcloud.com → 路径式 / +// 2. 虚拟主机式: mybucket-1250000000.cos.… → 路径为 / func bucketPathFor(endpoint, bucket string) string { if endpointHasBucket(endpoint, bucket) { return "/" diff --git a/frontend/src/api/system/index.ts b/frontend/src/api/system/index.ts index e816895..09c0211 100644 --- a/frontend/src/api/system/index.ts +++ b/frontend/src/api/system/index.ts @@ -51,6 +51,14 @@ export interface SelfCheckItem { detail?: string; } +/** 平台自身数据的保留期（天） */ +export interface RetentionSettings { + /** 告警发送历史保留天数，默认 7 */ + alertDeliveryDays: number + /** 审计日志保留天数，默认 90 */ + auditLogDays: number +} + export interface SelfCheckResult { checkedAt: string; overall: 'ok' | 'warn' | 'error'; @@ -99,4 +107,8 @@ export const systemApi = { }, restoreApply: () => http.post<{ message: string; dbPath: string }>('/system/restore/apply', { confirm: true }), + + // 平台数据保留期（告警发送历史 / 审计日志） + retention: () => http.get('/settings/retention'), + saveRetention: (data: RetentionSettings) => http.put('/settings/retention', data), }; diff --git a/frontend/src/i18n/locales/en/alerts.ts b/frontend/src/i18n/locales/en/alerts.ts index ea04391..50f7767 100644 --- a/frontend/src/i18n/locales/en/alerts.ts +++ b/frontend/src/i18n/locales/en/alerts.ts @@ -32,7 +32,7 @@ export default { lastFired: 'Last fired {{time}}', editAria: 'Edit {{name}}', deleteAria: 'Delete {{name}}', - historyTitle: 'Delivery history (last 50)', + historyTitle: 'Delivery history (kept {{days}} days)', notifyHintPre: 'Notifications not configured? Go to', notifyHintPost: 'to configure a Feishu (Lark) Webhook', editTitle: 'Edit rule · {{name}}', diff --git a/frontend/src/i18n/locales/en/analysis.ts b/frontend/src/i18n/locales/en/analysis.ts index 60388ce..13e412f 100644 --- a/frontend/src/i18n/locales/en/analysis.ts +++ b/frontend/src/i18n/locales/en/analysis.ts @@ -18,6 +18,7 @@ export default { namespaceEmpty: 'Loading namespaces…', namespaceAll: 'All namespaces (slow)', refresh: 'Refresh', + pendingApply: 'Filters changed — click Refresh to apply', }, stats: { total: 'Total logs', diff --git a/frontend/src/i18n/locales/en/common.ts b/frontend/src/i18n/locales/en/common.ts index 5a1ce1c..0a4d6ff 100644 --- a/frontend/src/i18n/locales/en/common.ts +++ b/frontend/src/i18n/locales/en/common.ts @@ -4,7 +4,7 @@ export default { menu: { groups: { resources: 'Resources', - services: 'Log Service', + services: 'Component Deploy', system: 'System Settings', }, dashboard: 'Dashboard', diff --git a/frontend/src/i18n/locales/en/settings.ts b/frontend/src/i18n/locales/en/settings.ts index 6eb0553..702b753 100644 --- a/frontend/src/i18n/locales/en/settings.ts +++ b/frontend/src/i18n/locales/en/settings.ts @@ -1,6 +1,18 @@ // settings namespace — Settings page (general/password/about, notifications, LDAP, image registry), English copy. // Note: theme & language card titles and ThemeSection copy are owned elsewhere and are not in this dictionary. export default { + retention: { + title: 'Log retention', + desc: 'How long the platform keeps its own records. Loki log retention itself is configured per deployment (Component Deploy → Retention).', + alertLabel: 'Alert delivery history (days)', + alertHint: 'Expired records are purged automatically; default 7 days', + auditLabel: 'Audit logs (days)', + auditHint: '90–180 days recommended for compliance', + save: 'Save', + saving: 'Saving…', + saved: 'Retention saved; expired records purged', + saveFailed: 'Save failed', + }, title: 'General', tabs: { general: 'General', diff --git a/frontend/src/i18n/locales/zh/alerts.ts b/frontend/src/i18n/locales/zh/alerts.ts index 576b4bb..3cc4d25 100644 --- a/frontend/src/i18n/locales/zh/alerts.ts +++ b/frontend/src/i18n/locales/zh/alerts.ts @@ -29,7 +29,7 @@ export default { lastFired: '上次触发 {{time}}', editAria: '编辑 {{name}}', deleteAria: '删除 {{name}}', - historyTitle: '发送历史(最近 50 条)', + historyTitle: '发送历史（保留 {{days}} 天）', notifyHintPre: '通知未配置?去', notifyHintPost: '配置飞书 Webhook', editTitle: '编辑规则 · {{name}}', diff --git a/frontend/src/i18n/locales/zh/analysis.ts b/frontend/src/i18n/locales/zh/analysis.ts index 325e77a..0a4c6e7 100644 --- a/frontend/src/i18n/locales/zh/analysis.ts +++ b/frontend/src/i18n/locales/zh/analysis.ts @@ -18,6 +18,7 @@ export default { namespaceEmpty: '正在加载命名空间…', namespaceAll: '全部命名空间（较慢）', refresh: '刷新', + pendingApply: '筛选条件已改，点「刷新」生效', }, stats: { total: '日志总量', diff --git a/frontend/src/i18n/locales/zh/common.ts b/frontend/src/i18n/locales/zh/common.ts index e0981b0..4366e23 100644 --- a/frontend/src/i18n/locales/zh/common.ts +++ b/frontend/src/i18n/locales/zh/common.ts @@ -4,7 +4,7 @@ export default { menu: { groups: { resources: '资源接入', - services: '日志服务', + services: '组件部署', system: '系统设置', }, dashboard: '仪表盘', diff --git a/frontend/src/i18n/locales/zh/settings.ts b/frontend/src/i18n/locales/zh/settings.ts index ce36e52..766c421 100644 --- a/frontend/src/i18n/locales/zh/settings.ts +++ b/frontend/src/i18n/locales/zh/settings.ts @@ -1,6 +1,18 @@ // settings 命名空间 — 系统设置页(通用配置/密码/关于、告警通知、LDAP 接入、镜像源)中文文案 // 注意:主题卡片与语言卡片的标题、ThemeSection 内文案由其他模块负责,不在本词典内。 export default { + retention: { + title: '日志保留时间', + desc: '平台自身数据的保留期。注意：Loki 日志本身的保留期在「组件部署 → 保留策略」里配置。', + alertLabel: '告警发送历史（天）', + alertHint: '过期记录自动清理，默认 7 天', + auditLabel: '审计日志（天）', + auditHint: '合规场景建议 90~180 天', + save: '保存', + saving: '保存中…', + saved: '保留策略已保存，过期数据已清理', + saveFailed: '保存失败', + }, title: '通用配置', tabs: { general: '通用配置', diff --git a/frontend/src/pages/alerts/index.tsx b/frontend/src/pages/alerts/index.tsx index bbb3715..a8ce3ca 100644 --- a/frontend/src/pages/alerts/index.tsx +++ b/frontend/src/pages/alerts/index.tsx @@ -21,6 +21,7 @@ import { EmptyState } from '@/components/EmptyState' import { ConfirmDialog } from '@/components/ConfirmDialog' import { alertsApi, type AlertRule, type AlertDelivery, type NotifyChannel } from '@/api/alerts' import { logsApi, type LogSource } from '@/api/logs' +import { systemApi } from '@/api/system' import { toast } from 'sonner' import { useForm } from 'react-hook-form' import { useTranslation } from 'react-i18next' @@ -85,6 +86,8 @@ const AlertsPage = memo(function AlertsPage() { const [sources, setSources] = useState([]) const [channels, setChannels] = useState([]) const [deliveries, setDeliveries] = useState([]) + /** 发送历史保留天数（系统设置 → 通用配置 → 日志保留时间），标题里展示 */ + const [historyDays, setHistoryDays] = useState(7) const [open, setOpen] = useState(false) const [editing, setEditing] = useState(null) const [deleteTarget, setDeleteTarget] = useState(null) @@ -99,9 +102,11 @@ const AlertsPage = memo(function AlertsPage() { const refresh = useCallback(async () => { try { - const [r, s, d, ch] = await Promise.all([ + const [r, s, d, ch, ret] = await Promise.all([ alertsApi.listRules(), logsApi.sources(), alertsApi.deliveries(), alertsApi.listChannels(), + systemApi.retention().catch(() => null), ]) + if (ret?.alertDeliveryDays) setHistoryDays(ret.alertDeliveryDays) setRules(Array.isArray(r) ? r : (r as unknown as { list: AlertRule[] }).list ?? []) setSources(s.list) setDeliveries(Array.isArray(d) ? d : []) @@ -308,7 +313,7 @@ const AlertsPage = memo(function AlertsPage() { ) : ( -

{t('historyTitle')}

+

{t('historyTitle', { days: historyDays })}

{deliveries.map((d) => (
diff --git a/frontend/src/pages/analysis/index.tsx b/frontend/src/pages/analysis/index.tsx index cd4df63..e31fea6 100644 --- a/frontend/src/pages/analysis/index.tsx +++ b/frontend/src/pages/analysis/index.tsx @@ -1,4 +1,4 @@ -import { memo, useEffect, useState } from 'react' +import { memo, useCallback, useEffect, useState } from 'react' import { useNavigate } from 'react-router' import { useTranslation } from 'react-i18next' import '@/i18n/domains' @@ -43,8 +43,16 @@ const AnalysisPage = memo(function AnalysisPage() { const [sourceId, setSourceId] = useState(0) // 默认 15 分钟：分析页一次性要跑 5 条 count_over_time， // 1 小时窗口在日志量大的命名空间要十几秒；15 分钟既能看出趋势又能快速出结果。 + // ── "选择值" 与 "已生效值" 分离 ────────────────────────────────── + // start/end/namespace 是**已生效**值，查询只认它们； + // sel* 是筛选栏里的**选择**值。 + // 这样切换命名空间/时间只改变选择，不触发查询；点「刷新」才生效。 + // 原因：一次分析要跑多条 count_over_time，切一下下拉就重查一轮很浪费， + // 用户快速连点几下会把 Loki 打满（并发护栏 429）。 const [start, setStart] = useState(() => toLocal(new Date(Date.now() - 15 * 60 * 1000))) const [end, setEnd] = useState(() => toLocal(new Date())) + const [selStart, setSelStart] = useState(() => toLocal(new Date(Date.now() - 15 * 60 * 1000))) + const [selEnd, setSelEnd] = useState(() => toLocal(new Date())) /** * 命名空间筛选。 @@ -56,6 +64,8 @@ const AnalysisPage = memo(function AnalysisPage() { // 初始为空，等策略返回后再定（不要用 localStorage 覆盖策略： // 策略是服务端下发的权威值，本地记忆会把管理员改过的策略"顶掉"）。 const [namespace, setNamespace] = useState('') + /** 命名空间下拉里的选择值（未点刷新前不参与查询） */ + const [selNamespace, setSelNamespace] = useState('') const [namespaceReady, setNamespaceReady] = useState(false) useEffect(() => { @@ -89,17 +99,24 @@ const AnalysisPage = memo(function AnalysisPage() { const values = (list || []).filter(Boolean).sort() const scopeNs = scope?.labels?.namespace ?? '' setNamespaces(values) - setNamespace(current => { + const resolve = (current: string) => { // 优先级：用户当前已选（同一数据源内不跳变）→ 授权策略 → default → 第一个 + // 注意：已移除「全部命名空间」选项 —— {namespace=~".+"} 要扫全集群 + // （实测约 370GB/天），一次分析跑几十秒，得不偿失。 const candidates = [current, scopeNs, 'default', values[0]] for (const candidate of candidates) { - if (!candidate) continue - if (candidate === ALL_NAMESPACES) return candidate + if (!candidate || candidate === ALL_NAMESPACES) continue if (values.includes(candidate)) return candidate } // 策略里配了但 Loki 里暂时没数据时，仍按策略显示（用户能看到"为什么没数据"） if (scopeNs) return scopeNs return '' + } + // 首次解析结果同时作为"选择值"和"已生效值"，保证进页面就有一屏数据 + setNamespace(current => { + const next = resolve(current) + setSelNamespace(next) + return next }) }) .catch(() => { if (!cancelled) setNamespaces([]) }) @@ -133,6 +150,20 @@ const AnalysisPage = memo(function AnalysisPage() { end, namespace: scopedNamespace, }) + + // 选择值与已生效值不一致 → 提示"点刷新生效" + const pendingApply = + namespaceReady && (selNamespace !== namespace || selStart !== start || selEnd !== end) + + /** 把筛选栏里的选择值应用到查询：点一次「刷新」才生效 */ + const applySelection = useCallback(() => { + setNamespace(selNamespace) + setStart(selStart) + setEnd(selEnd) + // 参数没变（只是点刷新）时也要重新取数 + refresh() + }, [selNamespace, selStart, selEnd, refresh]) + // 每日统计固定近 7 天,不随上方时间范围变化;错误就地展示,不影响主区 // 每日量可看范围：7 / 30 / 90 / 365 天（数据来自平台留存历史） const [dailyDays, setDailyDays] = useState(7) @@ -156,7 +187,7 @@ const AnalysisPage = memo(function AnalysisPage() { title={t('title')} description={t('subtitle')} action={ - @@ -179,9 +210,10 @@ const AnalysisPage = memo(function AnalysisPage() { - { setStart(nextStart); setEnd(nextEnd) }} /> + { setSelStart(nextStart); setSelEnd(nextEnd) }} /> + {pendingApply && ( + {t('filter.pendingApply')} + )}
{!namespaceReady && sourceId ? ( diff --git a/frontend/src/pages/settings/components/RetentionSection.tsx b/frontend/src/pages/settings/components/RetentionSection.tsx new file mode 100644 index 0000000..1bfb0ff --- /dev/null +++ b/frontend/src/pages/settings/components/RetentionSection.tsx @@ -0,0 +1,99 @@ +import { memo, useCallback, useEffect, useState } from 'react' +import { useTranslation } from 'react-i18next' +import { Button } from '@/components/ui/button' +import { Input } from '@/components/ui/input' +import { Card, CardContent, CardHeader, CardTitle } from '@/components/ui/card' +import { systemApi, type RetentionSettings } from '@/api/system' +import { toast } from 'sonner' + +/** + * 数据保留 — 平台自身产生的小数据保留多久。 + * + * 注意与「Loki 日志保留期」区分：那个是部署级配置（组件部署 → 保留策略）， + * 管的是日志本身；这里管的是平台库里的投递历史与审计日志。 + */ +export const RetentionSection = memo(function RetentionSection() { + const { t } = useTranslation('settings') + const [settings, setSettings] = useState({ alertDeliveryDays: 7, auditLogDays: 90 }) + const [loading, setLoading] = useState(true) + const [saving, setSaving] = useState(false) + + useEffect(() => { + let cancelled = false + systemApi + .retention() + .then(value => { + if (!cancelled && value) setSettings(value) + }) + .catch(() => undefined) + .finally(() => { + if (!cancelled) setLoading(false) + }) + return () => { + cancelled = true + } + }, []) + + const save = useCallback(async () => { + setSaving(true) + try { + const value = await systemApi.saveRetention(settings) + if (value) setSettings(value) + toast.success(t('retention.saved')) + } catch (err: unknown) { + toast.error(err instanceof Error ? err.message : t('retention.saveFailed')) + } finally { + setSaving(false) + } + }, [settings, t]) + + return ( + + + {t('retention.title')} +

{t('retention.desc')}

+ + +
+
+ + setSettings(s => ({ ...s, alertDeliveryDays: Number(e.target.value) || 0 }))} + /> +

{t('retention.alertHint')}

+
+
+ + setSettings(s => ({ ...s, auditLogDays: Number(e.target.value) || 0 }))} + /> +

{t('retention.auditHint')}

+
+
+
+ +
+ + + ) +}) diff --git a/frontend/src/pages/settings/index.tsx b/frontend/src/pages/settings/index.tsx index 1dea3c1..636eacc 100644 --- a/frontend/src/pages/settings/index.tsx +++ b/frontend/src/pages/settings/index.tsx @@ -11,6 +11,7 @@ import { clearAuth } from '@/lib/auth-storage' import { toast } from 'sonner' import { PageContainer } from '@/components/PageContainer' import { ThemeSection } from './components/ThemeSection' +import { RetentionSection } from './components/RetentionSection' import { LanguageSection } from '@/i18n/LanguageSection' import { Palette } from 'lucide-react' @@ -80,6 +81,7 @@ const SettingsPage = memo(function SettingsPage() { +