=== 0. 【严重】修复上一版引入的死锁(平台写路径全卡死)===
上一版把保留期读取写成 `GetRetentionSettings()`(内部 db.RLock()),
却在 `db.Lock()` 已持有的情况下调用(recordDelivery / RecordAudit 的裁剪路径)。
Go 的 sync.RWMutex **不可重入** → 第一次写投递历史/审计日志就永久死锁。
现象:登录 28 秒返回空(登录也要写审计)、所有写接口挂住。
go build / go vet / 之前的单测(不触发写路径)全都查不出来。
修法:
- 拆出 retentionFromDBLocked(db) / alertDeliveryRetentionLocked(db) /
auditLogRetentionLocked(db),调用方已持锁时用无锁版本
- upsertSettingLocked 里**不能**用 db.NextID()(内部 d.mu.Lock(),同样死锁),
改为锁内自算 max(id)+1
- 新增 retention_test.go:持写锁调用裁剪函数,2 秒超时未返回即判死锁
实测修复后:登录 103ms(原 28000ms)、panic/deadlock 命中 0。
=== 1. 平台数据保留期可配(系统设置 → 通用配置 → 日志保留时间)===
新增 GET/PUT /api/v1/settings/retention:
alertDeliveryDays 告警发送历史保留天数,默认 7(原写死"最近 50 条")
auditLogDays 审计日志保留天数,默认 90(原写死常量)
保存后立即按新策略裁剪一次,不必等下次触发。
校验:发送历史 1~365 天、审计 1~3650 天,超出返回明确中文原因(不是 500)。
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)。
=== 2. 发送历史:50 条硬上限 → 保留 7 天 ===
后端 Deliveries 接口原来写死 `len(out) < 50`,现在按保留期取(响应上限 500 可调),
标题改为「发送历史(保留 N 天)」,N 从设置里读。
=== 3. 菜单「日志服务」→「组件部署」===
=== 4. 分析页:去掉「全部命名空间」+ 改手动刷新 ===
- 移除「全部命名空间」选项:{namespace=~".+"} 要扫全集群(约 370GB/天),
一次分析几十秒,得不偿失;历史值 __all__ 也会被忽略并回退到策略/第一个 ns
- 筛选改为「选择值 / 已生效值」分离:切换命名空间或时间范围**不再自动查询**,
只把刷新按钮点亮并提示「筛选条件已改,点刷新生效」;
点一次才应用并取数。
原因:一次分析要跑多条 count_over_time,切一下下拉就重查一轮,
快速连点会把 Loki 打满(并发护栏 429),还拖慢别人。
=== 实测(9739c7c092e7)===
登录 103ms / token 273 字符
GET retention → {alertDeliveryDays:7, auditLogDays:90}
PUT 0 天 → 400「告警发送历史保留天数需在 1~365 之间」
平台日志 panic/deadlock 命中 0
go test ./... 全绿(含 2 个新死锁回归用例)
292 lines
8.2 KiB
Go
292 lines
8.2 KiB
Go
package controllers
|
|
|
|
import (
|
|
"strconv"
|
|
"time"
|
|
|
|
"lokistack/dto"
|
|
"lokistack/models"
|
|
"lokistack/repository"
|
|
"lokistack/services"
|
|
"lokistack/utils"
|
|
|
|
"github.com/gin-gonic/gin"
|
|
)
|
|
|
|
// AlertsController 告警规则 + 飞书通知配置
|
|
type AlertsController struct{}
|
|
|
|
func NewAlertsController() *AlertsController { return &AlertsController{} }
|
|
|
|
// ==================== 通知配置 ====================
|
|
|
|
// GetNotifyConfig 通知配置(只回 enabled/hasWebhook/hasSecret)
|
|
func (ctrl *AlertsController) GetNotifyConfig(c *gin.Context) {
|
|
cfg := services.GetNotifyConfig()
|
|
utils.Success(c, services.NotifyPublicConfig{
|
|
Enabled: cfg.Enabled, HasWebhook: cfg.Webhook != "", HasSecret: cfg.SigningSecret != "",
|
|
})
|
|
}
|
|
|
|
// UpdateNotifyConfig 保存(webhook/签名留空=保留旧值)
|
|
func (ctrl *AlertsController) UpdateNotifyConfig(c *gin.Context) {
|
|
var req dto.NotifyConfigRequest
|
|
if !utils.BindAndValidate(c, &req) {
|
|
return
|
|
}
|
|
if err := services.SaveNotifyConfig(req.Enabled, req.Webhook, req.SigningSecret); err != nil {
|
|
utils.Error(c, 400, err.Error())
|
|
return
|
|
}
|
|
utils.Success(c, gin.H{"message": "saved"})
|
|
}
|
|
|
|
// TestNotifyConfig 发送测试卡片(验证 webhook 连通)
|
|
func (ctrl *AlertsController) TestNotifyConfig(c *gin.Context) {
|
|
cfg := services.GetNotifyConfig()
|
|
if !cfg.Enabled || cfg.Webhook == "" {
|
|
utils.Error(c, 400, "请先保存并启用飞书通知")
|
|
return
|
|
}
|
|
if err := services.SendLarkCard("✅ LokiStack 通知测试", "green", []string{
|
|
"**时间**: " + time.Now().Format("2006-01-02 15:04:05"),
|
|
"**说明**: 管理员在设置页手动触发的连通性测试。",
|
|
}); err != nil {
|
|
utils.Error(c, 502, err.Error())
|
|
return
|
|
}
|
|
utils.Success(c, gin.H{"status": "sent"})
|
|
}
|
|
|
|
// ==================== 告警规则 CRUD ====================
|
|
|
|
func (ctrl *AlertsController) ListRules(c *gin.Context) {
|
|
db := repository.GetDB()
|
|
db.RLock()
|
|
out := make([]map[string]interface{}, 0, len(db.AlertRules))
|
|
for i := range db.AlertRules {
|
|
out = append(out, db.AlertRules[i].Sanitized())
|
|
}
|
|
db.RUnlock()
|
|
utils.Success(c, out)
|
|
}
|
|
|
|
func (ctrl *AlertsController) CreateRule(c *gin.Context) {
|
|
var req dto.AlertRuleRequest
|
|
if !utils.BindAndValidate(c, &req) {
|
|
return
|
|
}
|
|
if req.WindowMin <= 0 {
|
|
req.WindowMin = 15
|
|
}
|
|
if req.CooldownMin <= 0 {
|
|
req.CooldownMin = 60
|
|
}
|
|
if req.Threshold < 0 {
|
|
req.Threshold = 0
|
|
}
|
|
db := repository.GetDB()
|
|
rule := &models.AlertRule{
|
|
ID: db.NextID("alertRules"), Name: req.Name, SourceID: req.SourceID,
|
|
ChannelIDs: req.ChannelIDs,
|
|
Query: req.Query, WindowMin: req.WindowMin, Threshold: req.Threshold,
|
|
Enabled: req.Enabled, CooldownMin: req.CooldownMin,
|
|
}
|
|
db.Lock()
|
|
db.AlertRules = append(db.AlertRules, *rule)
|
|
db.RequestSave()
|
|
db.Unlock()
|
|
utils.Success(c, rule.Sanitized())
|
|
}
|
|
|
|
func (ctrl *AlertsController) UpdateRule(c *gin.Context) {
|
|
id, err := strconv.ParseUint(c.Param("id"), 10, 64)
|
|
if err != nil {
|
|
utils.BadRequest(c, "ID 无效")
|
|
return
|
|
}
|
|
var req map[string]interface{}
|
|
if !utils.BindAndValidate(c, &req) {
|
|
return
|
|
}
|
|
db := repository.GetDB()
|
|
db.Lock()
|
|
defer db.Unlock()
|
|
for i := range db.AlertRules {
|
|
if db.AlertRules[i].ID != uint(id) {
|
|
continue
|
|
}
|
|
if v, ok := req["name"].(string); ok && v != "" {
|
|
db.AlertRules[i].Name = v
|
|
}
|
|
if v, ok := req["query"].(string); ok && v != "" {
|
|
db.AlertRules[i].Query = v
|
|
}
|
|
if v, ok := req["sourceId"].(float64); ok && v > 0 {
|
|
db.AlertRules[i].SourceID = uint(v)
|
|
}
|
|
if arr, ok := req["channelIds"].([]interface{}); ok {
|
|
ids := make([]uint, 0, len(arr))
|
|
for _, item := range arr {
|
|
if v, ok := item.(float64); ok && v > 0 {
|
|
ids = append(ids, uint(v))
|
|
}
|
|
}
|
|
db.AlertRules[i].ChannelIDs = ids
|
|
}
|
|
if v, ok := req["windowMin"].(float64); ok && v > 0 {
|
|
db.AlertRules[i].WindowMin = int(v)
|
|
}
|
|
if v, ok := req["threshold"].(float64); ok && v >= 0 {
|
|
db.AlertRules[i].Threshold = int(v)
|
|
}
|
|
if v, ok := req["cooldownMin"].(float64); ok && v > 0 {
|
|
db.AlertRules[i].CooldownMin = int(v)
|
|
}
|
|
if v, ok := req["enabled"].(bool); ok {
|
|
db.AlertRules[i].Enabled = v
|
|
}
|
|
db.AlertRules[i].UpdatedAt = time.Now()
|
|
db.RequestSave()
|
|
utils.Success(c, db.AlertRules[i].Sanitized())
|
|
return
|
|
}
|
|
utils.NotFound(c, "规则不存在")
|
|
}
|
|
|
|
func (ctrl *AlertsController) DeleteRule(c *gin.Context) {
|
|
id, err := strconv.ParseUint(c.Param("id"), 10, 64)
|
|
if err != nil {
|
|
utils.BadRequest(c, "ID 无效")
|
|
return
|
|
}
|
|
db := repository.GetDB()
|
|
db.Lock()
|
|
defer db.Unlock()
|
|
for i := range db.AlertRules {
|
|
if db.AlertRules[i].ID == uint(id) {
|
|
db.AlertRules = append(db.AlertRules[:i], db.AlertRules[i+1:]...)
|
|
db.RequestSave()
|
|
utils.Success(c, gin.H{"message": "deleted"})
|
|
return
|
|
}
|
|
}
|
|
utils.NotFound(c, "规则不存在")
|
|
}
|
|
|
|
// CheckNow 手动触发一次全量检查(不等定时器)
|
|
func (ctrl *AlertsController) CheckNow(c *gin.Context) {
|
|
services.RunAlertCheck(c.Request.Context())
|
|
utils.Success(c, gin.H{"message": "checked"})
|
|
}
|
|
|
|
// 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, 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,
|
|
"count": d.Count, "status": d.Status, "reason": d.Reason, "createdAt": d.CreatedAt,
|
|
})
|
|
}
|
|
db.RUnlock()
|
|
utils.Success(c, out)
|
|
}
|
|
|
|
// ==================== 通知通道(多 webhook) ====================
|
|
|
|
func (ctrl *AlertsController) ListChannels(c *gin.Context) {
|
|
utils.Success(c, services.ListNotifyChannels())
|
|
}
|
|
|
|
func (ctrl *AlertsController) SaveChannel(c *gin.Context) {
|
|
var req dto.NotifyChannelRequest
|
|
if !utils.BindAndValidate(c, &req) {
|
|
return
|
|
}
|
|
// PUT /channels/:id 时路径参数优先
|
|
if pathID := c.Param("id"); pathID != "" {
|
|
if v, err := strconv.ParseUint(pathID, 10, 64); err == nil {
|
|
req.ID = uint(v)
|
|
}
|
|
}
|
|
ch, err := services.SaveNotifyChannel(req.ID, req.Name, req.Webhook, req.SigningSecret, req.Enabled)
|
|
if err != nil {
|
|
utils.Error(c, 400, err.Error())
|
|
return
|
|
}
|
|
utils.Success(c, ch.Sanitized())
|
|
}
|
|
|
|
func (ctrl *AlertsController) DeleteChannel(c *gin.Context) {
|
|
id, err := strconv.ParseUint(c.Param("id"), 10, 64)
|
|
if err != nil {
|
|
utils.BadRequest(c, "ID 无效")
|
|
return
|
|
}
|
|
if err := services.DeleteNotifyChannel(uint(id)); err != nil {
|
|
utils.Error(c, 400, err.Error())
|
|
return
|
|
}
|
|
utils.Success(c, gin.H{"message": "deleted"})
|
|
}
|
|
|
|
// TestChannel 向指定通道发送测试卡片(用表单参数或已存通道)
|
|
func (ctrl *AlertsController) TestChannel(c *gin.Context) {
|
|
var req dto.NotifyChannelRequest
|
|
if !utils.BindAndValidate(c, &req) {
|
|
return
|
|
}
|
|
// 测试已存通道时 name 可不传(沿用原名称);直测新 webhook 时缺省名
|
|
name := req.Name
|
|
if req.ID > 0 {
|
|
if existing := services.GetNotifyChannel(req.ID); existing != nil && name == "" {
|
|
name = existing.Name
|
|
}
|
|
} else if name == "" {
|
|
name = "临时测试通道"
|
|
}
|
|
var channelID uint
|
|
if req.ID > 0 {
|
|
channelID = req.ID
|
|
// 表单 webhook 非空时先保存再测(未保存也可测新地址)
|
|
if req.Webhook != "" {
|
|
if _, err := services.SaveNotifyChannel(req.ID, name, req.Webhook, req.SigningSecret, true); err != nil {
|
|
utils.Error(c, 400, err.Error())
|
|
return
|
|
}
|
|
}
|
|
} else if req.Webhook != "" {
|
|
// 未保存直测:保存后可删
|
|
ch, err := services.SaveNotifyChannel(0, name, req.Webhook, req.SigningSecret, true)
|
|
if err != nil {
|
|
utils.Error(c, 400, err.Error())
|
|
return
|
|
}
|
|
channelID = ch.ID
|
|
}
|
|
if channelID == 0 {
|
|
utils.BadRequest(c, "请提供通道 ID 或 webhook 地址")
|
|
return
|
|
}
|
|
if err := services.SendLarkCardToChannel(channelID, "✅ LokiStack 通道测试", "green", []string{
|
|
"**时间**: " + time.Now().Format("2006-01-02 15:04:05"),
|
|
"**说明**: 通道连通性测试。",
|
|
}); err != nil {
|
|
utils.Error(c, 502, err.Error())
|
|
return
|
|
}
|
|
utils.Success(c, gin.H{"status": "sent"})
|
|
}
|