Files
LokiStack/backend/routers/router.go
T
opsx 5e59347c40 feat(settings,analysis): 日志保留时间可配(发送历史保留 7 天)+ 分析页手动刷新;修复我引入的 RWMutex 死锁
=== 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)
2026-10-29 04:16:52 +00:00

291 lines
12 KiB
Go
Raw Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
package routers
import (
"net/http"
"os"
"path/filepath"
"strings"
"time"
"lokistack/controllers"
"lokistack/middleware"
"lokistack/pkg/metrics"
"github.com/gin-contrib/cors"
"github.com/gin-gonic/gin"
)
func SetupRouter() *gin.Engine {
// 用 gin.New() 而不是 gin.Default():Default 自带文本格式的 gin.Logger(),
// 这里换成结构化 JSON 访问日志(middleware.AccessLog),便于 Promtail→Loki
// 按字段过滤,同时跳过健康检查/静态资源,避免探针轮询把日志量刷满。
// Recovery 必须显式加回来(Default 里包含它),否则 panic 会打挂进程。
r := gin.New()
// AuditLog 放在最外层:它依赖的 userId/username 由 group 内的 JWTAuth 写入,
// 但中间件的后置逻辑在 c.Next() 返回后才执行,那时已就绪(见 middleware/audit.go)。
r.Use(middleware.AccessLog(), middleware.AuditLog(), middleware.Metrics(), gin.Recovery())
// CORS
cfg := cors.DefaultConfig()
cfg.AllowAllOrigins = true
cfg.AllowHeaders = []string{"Origin", "Content-Type", "Accept", "Authorization", "X-Requested-With"}
cfg.AllowMethods = []string{"GET", "POST", "PUT", "DELETE", "OPTIONS", "PATCH"}
r.Use(cors.New(cfg))
// Controllers
authCtrl := controllers.NewAuthController()
clusterCtrl := controllers.NewClusterController()
hostCtrl := controllers.NewHostController()
storageCtrl := controllers.NewStorageController()
deploymentCtrl := controllers.NewDeploymentController()
collectorCtrl := controllers.NewCollectorController()
userCtrl := controllers.NewUserController()
ldapCtrl := controllers.NewLdapController()
logsCtrl := controllers.NewLogsController()
alertsCtrl := controllers.NewAlertsController()
auditCtrl := controllers.NewAuditController()
systemCtrl := controllers.NewSystemController()
registryCtrl := controllers.NewImageRegistryController()
// API v1
api := r.Group("/api/v1")
// 健康检查(公开访问,用于负载均衡和监控)
api.GET("/health", func(c *gin.Context) {
c.JSON(http.StatusOK, gin.H{
"status": "ok",
"timestamp": time.Now().Unix(),
"version": "2.0.0",
})
})
// Prometheus metrics 端点(真实指标,见 services/metrics.go)
// 原实现只返回一行注释却声称"可被 Prometheus 抓取",平台自身等于不可观测。
api.GET("/metrics", gin.WrapH(metrics.Handler()))
// Public(登录接口挂 IP 限流防爆破:20 次/分钟)
api.POST("/auth/login", middleware.LoginRateLimit(20, time.Minute), authCtrl.Login)
api.POST("/auth/refresh", authCtrl.RefreshToken)
// Auth required
authorized := api.Group("")
authorized.Use(middleware.JWTAuth())
// Auth routes
authorized.GET("/auth/user", authCtrl.GetUser)
authorized.POST("/auth/password", authCtrl.ChangePassword)
// 日志查询:数据源管理 + Loki 代理查询
logs := authorized.Group("/logs")
{
logs.GET("/sources", logsCtrl.ListSources)
logs.POST("/sources", middleware.AdminAuth(), logsCtrl.CreateSource)
logs.PUT("/sources/:id", middleware.AdminAuth(), logsCtrl.UpdateSource)
logs.DELETE("/sources/:id", middleware.AdminAuth(), logsCtrl.DeleteSource)
logs.POST("/sources/test", middleware.AdminAuth(), logsCtrl.TestSource)
logs.GET("/policy-groups", middleware.AdminAuth(), logsCtrl.ListPolicyGroups)
logs.POST("/policy-groups", middleware.AdminAuth(), logsCtrl.CreatePolicyGroup)
logs.PUT("/policy-groups/:id", middleware.AdminAuth(), logsCtrl.UpdatePolicyGroup)
logs.DELETE("/policy-groups/:id", middleware.AdminAuth(), logsCtrl.DeletePolicyGroup)
logs.POST("/query", logsCtrl.Query)
logs.POST("/context", logsCtrl.Context)
logs.GET("/labels", logsCtrl.Labels)
logs.GET("/label/:name/values", logsCtrl.LabelValues)
logs.GET("/my-scope", logsCtrl.MyScope)
// ⚠️ 必须是 POST:前端 logsApi.histogram 用 POST 传 body(sourceId/query/start/end)。
// 曾被误改成 GET,导致 POST /logs/histogram 全部 404,
// 表现为日志查询页直方图不显示、分析页直接报 "API endpoint not found"。
// 同时保留 GET 作为简单调试入口(参数放 query string)。
logs.POST("/histogram", logsCtrl.Histogram)
logs.GET("/histogram", logsCtrl.Histogram)
logs.GET("/daily-volume", logsCtrl.DailyVolume)
// 总量统计(index/stats):只读索引,用于算总页数
logs.POST("/stats", logsCtrl.Stats)
logs.POST("/metrics", middleware.AdminAuth(), logsCtrl.Metrics)
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)
// 告警:飞书通知配置 + 告警规则
alertNotify := authorized.Group("/alerts")
{
alertNotify.GET("/notify-config", alertsCtrl.GetNotifyConfig)
alertNotify.PUT("/notify-config", middleware.AdminAuth(), alertsCtrl.UpdateNotifyConfig)
alertNotify.POST("/notify-config/test", middleware.AdminAuth(), alertsCtrl.TestNotifyConfig)
alertNotify.GET("/rules", alertsCtrl.ListRules)
alertNotify.POST("/rules", middleware.AdminAuth(), alertsCtrl.CreateRule)
alertNotify.PUT("/rules/:id", middleware.AdminAuth(), alertsCtrl.UpdateRule)
alertNotify.DELETE("/rules/:id", middleware.AdminAuth(), alertsCtrl.DeleteRule)
alertNotify.POST("/check", middleware.AdminAuth(), alertsCtrl.CheckNow)
alertNotify.GET("/deliveries", alertsCtrl.Deliveries)
alertNotify.GET("/channels", alertsCtrl.ListChannels)
alertNotify.POST("/channels", middleware.AdminAuth(), alertsCtrl.SaveChannel)
alertNotify.PUT("/channels/:id", middleware.AdminAuth(), alertsCtrl.SaveChannel)
alertNotify.DELETE("/channels/:id", middleware.AdminAuth(), alertsCtrl.DeleteChannel)
alertNotify.POST("/channels/test", middleware.AdminAuth(), alertsCtrl.TestChannel)
}
// LDAP 配置(admin 读写;查询供用户页判断导入按钮显隐,登录用户即可读)
authorized.GET("/settings/ldap", middleware.AdminAuth(), ldapCtrl.GetConfig)
authorized.PUT("/settings/ldap", middleware.AdminAuth(), ldapCtrl.UpdateConfig)
authorized.POST("/settings/ldap/test", middleware.AdminAuth(), ldapCtrl.TestConnection)
authorized.POST("/settings/ldap/users", middleware.AdminAuth(), ldapCtrl.ListUsers)
// Users(admin 专用)
// 系统运维:备份 / 恢复 / 一键自检(仅管理员)
system := authorized.Group("/system")
system.Use(middleware.AdminAuth())
{
system.GET("/backup", systemCtrl.Backup)
system.POST("/restore", systemCtrl.Restore)
system.POST("/restore/apply", systemCtrl.RestoreApply)
system.GET("/backups", systemCtrl.Backups)
system.GET("/self-check", systemCtrl.Health)
}
// 审计日志:仅管理员可查(含他人操作记录)
audit := authorized.Group("/audit-logs")
audit.Use(middleware.AdminAuth())
{
audit.GET("", auditCtrl.List)
audit.GET("/stats", auditCtrl.Stats)
audit.GET("/:id", auditCtrl.Get)
}
users := authorized.Group("/users")
users.Use(middleware.AdminAuth())
{
users.GET("", userCtrl.List)
users.POST("", userCtrl.Create)
users.DELETE("/:id", userCtrl.Delete)
users.POST("/batch-delete", userCtrl.BatchDelete)
users.POST("/:id/reset-password", userCtrl.ResetPassword)
users.POST("/batch-import-ldap", userCtrl.BatchImportLDAP)
}
// 普通用户仅可访问日志查询及自己的账号接口;管理资源统一在服务端保护。
authorized.Use(middleware.AdminAuth())
// Clusters
authorized.GET("/clusters", clusterCtrl.List)
authorized.GET("/clusters/:id", clusterCtrl.Get)
authorized.POST("/clusters", clusterCtrl.Create)
authorized.PUT("/clusters/:id", clusterCtrl.Update)
authorized.DELETE("/clusters/:id", clusterCtrl.Delete)
authorized.POST("/clusters/:id/validate", clusterCtrl.Validate)
authorized.GET("/clusters/:id/info", clusterCtrl.GetInfo)
authorized.GET("/clusters/:id/namespaces", clusterCtrl.Namespaces)
authorized.GET("/clusters/:id/namespaces/:namespace/deployments", clusterCtrl.Deployments)
// Hosts
authorized.GET("/hosts", hostCtrl.List)
authorized.GET("/hosts/:id", hostCtrl.Get)
authorized.POST("/hosts", hostCtrl.Create)
authorized.PUT("/hosts/:id", hostCtrl.Update)
authorized.DELETE("/hosts/:id", hostCtrl.Delete)
authorized.POST("/hosts/:id/validate", hostCtrl.Validate)
authorized.GET("/hosts/:id/info", hostCtrl.GetInfo)
// Storage
authorized.GET("/storages", storageCtrl.List)
authorized.GET("/storages/:id", storageCtrl.Get)
authorized.POST("/storages", storageCtrl.Create)
authorized.POST("/storages/test", storageCtrl.TestConnection)
authorized.PUT("/storages/:id", storageCtrl.Update)
authorized.DELETE("/storages/:id", storageCtrl.Delete)
authorized.GET("/storages/default", storageCtrl.GetDefault)
// Deployments
authorized.GET("/deployments", deploymentCtrl.List)
authorized.GET("/deployments/:id", deploymentCtrl.Get)
authorized.POST("/deployments", deploymentCtrl.Create)
authorized.PUT("/deployments/:id", deploymentCtrl.Update)
authorized.POST("/deployments/:id/retention", deploymentCtrl.UpdateRetention)
authorized.GET("/deployments/:id/config", middleware.AdminAuth(), deploymentCtrl.GetLiveConfig)
authorized.DELETE("/deployments/:id", deploymentCtrl.Delete)
authorized.POST("/deployments/:id/deploy", deploymentCtrl.Deploy)
authorized.POST("/deployments/:id/stop", deploymentCtrl.Stop)
authorized.POST("/deployments/:id/restart", deploymentCtrl.Restart)
authorized.GET("/deployments/:id/logs", deploymentCtrl.GetLogs)
// 资源用量(CPU/内存/限流),页面用来判断是否要扩容
authorized.GET("/deployments/:id/usage", deploymentCtrl.GetUsage)
authorized.GET("/deployments/stats", deploymentCtrl.GetStats)
authorized.GET("/deployments/preflight", deploymentCtrl.Preflight)
authorized.GET("/deployments/:id/ready", deploymentCtrl.Ready)
authorized.GET("/deployments/versions", deploymentCtrl.GetVersions)
// Collectors
authorized.GET("/collectors", collectorCtrl.List)
authorized.GET("/collectors/:id", collectorCtrl.Get)
authorized.POST("/collectors", collectorCtrl.Create)
authorized.PUT("/collectors/:id", collectorCtrl.Update)
authorized.DELETE("/collectors/:id", collectorCtrl.Delete)
authorized.POST("/collectors/:id/install", collectorCtrl.Install)
authorized.POST("/collectors/:id/uninstall", collectorCtrl.Uninstall)
authorized.POST("/collectors/:id/preview", collectorCtrl.Preview)
authorized.GET("/collectors/:id/logs", collectorCtrl.Logs)
// Static files - use executable directory as base
distPath := findDistPath()
if distPath != "" {
r.Static("/assets", filepath.Join(distPath, "assets"))
// favicon:NoRoute 会把所有非 /api/ 请求回退到 index.html,
// 故此处显式注册 /favicon.svg 与 /favicon.ico,否则浏览器请求图标只会拿到 HTML。
faviconSVG := filepath.Join(distPath, "favicon.svg")
serveFavicon := func(c *gin.Context) {
if _, err := os.Stat(faviconSVG); err != nil {
c.Status(404)
return
}
c.Header("Content-Type", "image/svg+xml")
c.File(faviconSVG)
}
r.GET("/favicon.svg", serveFavicon)
r.GET("/favicon.ico", serveFavicon)
r.NoRoute(func(c *gin.Context) {
if strings.HasPrefix(c.Request.URL.Path, "/api/") {
c.JSON(404, gin.H{"code": 404, "message": "API endpoint not found"})
return
}
c.File(filepath.Join(distPath, "index.html"))
})
}
return r
}
// findDistPath finds the frontend dist directory
func findDistPath() string {
// Try relative to executable
if ex, err := os.Executable(); err == nil {
execDir := filepath.Dir(ex)
// If running via "go run", executable is in temp dir - fall through
if !strings.Contains(execDir, "/tmp/go-build") {
for _, rel := range []string{"dist", "../dist", "../../dist"} {
p := filepath.Join(execDir, rel)
if _, err := os.Stat(filepath.Join(p, "index.html")); err == nil {
return p
}
}
}
}
// Try working directory relative paths
for _, p := range []string{"./dist", "../dist", "/app/dist"} {
if abs, err := filepath.Abs(p); err == nil {
if _, err := os.Stat(filepath.Join(abs, "index.html")); err == nil {
return abs
}
}
}
return ""
}