Files
LokiStack/backend/services/log_source.go
T
opsx 95dcd9408c fix(logs,analysis,alerts): 数据源不可达、分析页超时、告警规则混入基础设施项
=== 1. 主机模式数据源地址 0.0.0.0 → 主机真实 IP(查询页"永远加载中"的根因)===
现象:切换数据源后命名空间/App 下拉不动,一直显示"加载中…",控制台报
    Get "http://0.0.0.0:3100/loki/api/v1/labels": dial tcp 0.0.0.0:3100: connect: connection refused

根因:主机模式部署的 endpoint 记的是**容器监听地址** 0.0.0.0:3100,
而平台进程不在被采集主机上,0.0.0.0 只在本机有意义。
EnsureLogSourceForDeployment 把它原样写成了数据源 URL。

修法:
  - 新增 reachableEndpoint():把 0.0.0.0 / [::] 换成关联主机的真实 IP
  - schemaVersion v3→v4 迁移 fixUnreachableLogSourceURLs(),修正已有脏数据
  - 实测:http://0.0.0.0:3100 → http://172.19.16.8:3100,该源 label 查询由 502 变 200

=== 2. 前端:区分"加载中"与"加载失败" ===
原来用 `loading={!options.length}` 表示加载中 —— 数据源不可达时列表永远为空,
于是界面永远转圈,把真实错误完全藏起来了(这次的 0.0.0.0 就是这样被掩盖的)。
改为独立的 namespaceLoading / namespaceError 两个状态:
  - 切换数据源先清空旧 options(否则显示上一个源的命名空间,看着像"没反应")
  - 失败显示"加载失败"并在旁边给「重试」按钮,不用刷新整页

=== 3. 日志分析页:主图改走索引接口(原来 5 条 count_over_time 全超时)===
实测 astralx(约 370GB/天)15 分钟窗口:
  count_over_time   → exec 29s + 排队 95s,超过 30s 护栏 → 502
  index/volume_range → 0s
改为两阶段渲染:
  - 总量趋势改用 /logs/histogram(内部 index/volume_range),**0s 出图**
  - 错误/警告/app 维度仍需读日志内容,串行执行,先渲染主图再更新
  - heavyQueryTimeout 30s → 90s(实测 error 维度查询 14s 可完成)

=== 4. 告警规则按日志平台定位收敛 ===
  - seed 里删掉「Kubernetes 节点 NotReady」「容器 OOMKilled」:
    这类属于 kube-state-metrics / Prometheus 的职责,用日志字符串匹配不可靠
  - 迁移里清掉线上已存在的同类规则(3 条)
  - 修 AlertRule / NotifyChannel 重复 ID(同 logSources 一样漏在 sequences 重建外)

=== 5. 告警页 UI ===
  - 规则列表由"每条一张卡"改为紧凑表格(一屏看全,阈值可对齐比较)
  - 通知通道(飞书/Lark/Webhook)从「系统设置」搬到告警页折叠区:
    规则和"发到哪"是一件事的两半,分两页配容易漏配导致告警发不出去

=== 6. 权限收敛 ===
  - /alerts 加 adminOnly:普通用户按查询授权只看「日志查询」,不再看到告警菜单
  - 系统设置从侧栏移到右上角齿轮(低频平台配置,不占侧栏一位)

=== 7. 迁移设计教训(写进代码注释)===
v2 时把"告警去重 + 清基础设施规则"塞进 case 1,但线上库已到 v2,
那段代码永远不执行。**结构变更只要逻辑变了、目标库已达该版本,就必须再抬一版**,
不能改历史迁移分支。

=== 实测(v4-1028-2210 / 979bdb4a48e8)===
  v4 迁移:数据源 URL 修正 http://0.0.0.0:3100 → http://172.19.16.8:3100
  source 1 label:200,0s(astralx / log)
  source 2 label:200,0s(不再 connection refused)
  分析页 histogram astralx 15m:0s
  分析页 metrics  error  astralx 15m:14s
  告警规则:10 条(已无基础设施类)、ID 唯一
  另清理 172.19.16.8 上 13 个残留容器,回收 6.1GB
2026-10-28 22:11:48 +00:00

451 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.
// 日志数据源管理 — CRUD、默认源、部署自动注册(Managed 标记)。
package services
import (
"encoding/json"
"fmt"
"regexp"
"sort"
"strings"
"time"
"lokistack/models"
"lokistack/pkg/crypto"
"lokistack/repository"
)
// ==================== 数据源 CRUD ====================
func GetLogSource(id uint) *models.LogSource {
db := repository.GetDB()
db.RLock()
defer db.RUnlock()
for i := range db.LogSources {
if db.LogSources[i].ID == id {
s := db.LogSources[i]
return &s
}
}
return nil
}
func ListLogSources() []map[string]interface{} {
db := repository.GetDB()
db.RLock()
defer db.RUnlock()
out := make([]map[string]interface{}, 0, len(db.LogSources))
for i := range db.LogSources {
out = append(out, db.LogSources[i].Sanitized())
}
return out
}
func ListVisibleLogSources(userID uint, admin bool) []map[string]interface{} {
db := repository.GetDB()
db.RLock()
defer db.RUnlock()
out := make([]map[string]interface{}, 0)
for i := range db.LogSources {
s := &db.LogSources[i]
if admin || policyGroupAccess(db.LogPolicyGroups, s.ID, userID) != nil || sourceAccess(s, userID) != nil {
item := s.Sanitized()
if !admin {
item = map[string]interface{}{"id": s.ID, "name": s.Name, "isDefault": s.IsDefault}
}
out = append(out, item)
}
}
return out
}
// UserAccessLabels 返回某用户在指定数据源上的授权 labels。
//
// 返回值语义:
// - nil → 该用户对此数据源完全无权限
// - 空 map → 有权限但不限定范围(可查全部)
// - 非空 map → 只能查匹配这些 label 的日志(如 {"namespace":"astralx"})
func UserAccessLabels(src *models.LogSource, userID uint) map[string]string {
db := repository.GetDB()
var access *models.LogSourceAccess
if db != nil {
db.RLock()
access = policyGroupAccess(db.LogPolicyGroups, src.ID, userID)
db.RUnlock()
}
if access == nil {
access = sourceAccess(src, userID)
}
if access == nil {
return nil
}
if access.Labels == nil {
return map[string]string{}
}
return access.Labels
}
func sourceAccess(src *models.LogSource, userID uint) *models.LogSourceAccess {
for i := range src.Access {
if src.Access[i].UserID == userID {
return &src.Access[i]
}
}
return nil
}
var labelNamePattern = regexp.MustCompile(`^[a-zA-Z_][a-zA-Z0-9_]*$`)
func ValidLogLabelName(name string) bool { return labelNamePattern.MatchString(name) }
// AuthorizedLogQuery constrains the Loki stream selector server-side. Non-admins must be explicitly granted access.
func AuthorizedLogQuery(src *models.LogSource, userID uint, admin bool, query string) (string, error) {
if admin {
return query, nil
}
db := repository.GetDB()
var access *models.LogSourceAccess
if db != nil {
db.RLock()
access = policyGroupAccess(db.LogPolicyGroups, src.ID, userID)
db.RUnlock()
}
if access == nil {
access = sourceAccess(src, userID)
}
if access == nil {
return "", fmt.Errorf("无权查询此数据源")
}
q := strings.TrimSpace(query)
if !strings.HasPrefix(q, "{") {
return "", fmt.Errorf("普通用户查询必须以日志流选择器开头")
}
close := strings.IndexByte(q, '}')
if close < 0 || strings.Contains(q[close+1:], "{") {
return "", fmt.Errorf("查询只能包含一个日志流选择器")
}
if len(access.Labels) == 0 {
return q, nil
}
keys := make([]string, 0, len(access.Labels))
for k := range access.Labels {
if !labelNamePattern.MatchString(k) {
return "", fmt.Errorf("访问策略包含无效标签名")
}
keys = append(keys, k)
}
sort.Strings(keys)
matchers := make([]string, 0, len(keys))
for _, k := range keys {
v, _ := json.Marshal(access.Labels[k])
matchers = append(matchers, k+"="+string(v))
}
inner := strings.TrimSpace(q[1:close])
if inner != "" {
inner += ","
}
return "{" + inner + strings.Join(matchers, ",") + "}" + q[close+1:], nil
}
func CreateLogSource(s *models.LogSource) (*models.LogSource, error) {
if err := encryptSourceCredentials(s); err != nil {
return nil, err
}
db := repository.GetDB()
newID := db.NextID("logSources")
db.Lock()
defer db.Unlock()
s.ID = newID
if s.IsDefault {
for i := range db.LogSources {
db.LogSources[i].IsDefault = false
}
}
// 首个数据源自动设默认
if len(db.LogSources) == 0 {
s.IsDefault = true
}
db.LogSources = append(db.LogSources, *s)
db.RequestSave()
return s, nil
}
func encryptSourceCredentials(s *models.LogSource) error {
if s.BasicPass != "" {
v, err := crypto.EncryptField(s.BasicPass)
if err != nil {
return err
}
s.BasicPass = v
}
if s.BearerToken != "" {
v, err := crypto.EncryptField(s.BearerToken)
if err != nil {
return err
}
s.BearerToken = v
}
for name, value := range s.Headers {
if !validSourceHeader(name) {
return fmt.Errorf("自定义 Header 名称无效: %s", name)
}
v, err := crypto.EncryptField(value)
if err != nil {
return err
}
s.Headers[name] = v
}
return nil
}
func validSourceHeader(name string) bool {
if name == "" || strings.EqualFold(name, "Host") || strings.EqualFold(name, "Authorization") {
return false
}
for _, c := range name {
if !(c >= 'a' && c <= 'z' || c >= 'A' && c <= 'Z' || c >= '0' && c <= '9' || c == '-') {
return false
}
}
return true
}
func UpdateLogSource(id uint, patch map[string]interface{}) error {
db := repository.GetDB()
db.Lock()
defer db.Unlock()
for i := range db.LogSources {
if db.LogSources[i].ID != id {
continue
}
if v, ok := patch["name"].(string); ok && v != "" {
db.LogSources[i].Name = v
}
if v, ok := patch["namespace"].(string); ok {
db.LogSources[i].Namespace = v
}
if v, ok := patch["service"].(string); ok {
db.LogSources[i].Service = v
}
if v, ok := patch["port"].(float64); ok {
db.LogSources[i].Port = int(v)
}
if v, ok := patch["url"].(string); ok {
db.LogSources[i].URL = v
}
if v, ok := patch["authType"].(string); ok {
db.LogSources[i].AuthType = v
}
if v, ok := patch["basicUser"].(string); ok {
db.LogSources[i].BasicUser = v
}
if v, ok := patch["bearerToken"].(string); ok {
if v != "" {
enc, err := crypto.EncryptField(v)
if err != nil {
return err
}
db.LogSources[i].BearerToken = enc
}
}
if v, ok := patch["skipTLSVerify"].(bool); ok {
db.LogSources[i].SkipTLSVerify = v
}
if raw, ok := patch["headers"]; ok {
data, err := json.Marshal(raw)
if err != nil {
return err
}
var headers map[string]string
if err := json.Unmarshal(data, &headers); err != nil {
return err
}
for name, value := range headers {
if !validSourceHeader(name) {
return fmt.Errorf("自定义 Header 名称无效: %s", name)
}
if value == "" {
if old, ok := db.LogSources[i].Headers[name]; ok {
headers[name] = old
}
continue
}
enc, err := crypto.EncryptField(value)
if err != nil {
return err
}
headers[name] = enc
}
db.LogSources[i].Headers = headers
}
if v, ok := patch["basicPass"].(string); ok && v != "" {
if enc, err := crypto.EncryptField(v); err == nil {
db.LogSources[i].BasicPass = enc
}
}
if v, ok := patch["isDefault"].(bool); ok && v {
for j := range db.LogSources {
db.LogSources[j].IsDefault = db.LogSources[j].ID == id
}
}
if raw, ok := patch["access"]; ok {
data, err := json.Marshal(raw)
if err != nil {
return err
}
var access []models.LogSourceAccess
if err := json.Unmarshal(data, &access); err != nil {
return err
}
if err := ValidateSourceAccess(access); err != nil {
return err
}
db.LogSources[i].Access = access
}
db.LogSources[i].UpdatedAt = time.Now()
db.RequestSave()
return nil
}
return fmt.Errorf("数据源不存在")
}
func ValidateSourceAccess(access []models.LogSourceAccess) error {
seen := map[uint]bool{}
for _, rule := range access {
if rule.UserID == 0 || seen[rule.UserID] {
return fmt.Errorf("访问策略的用户 ID 无效或重复")
}
seen[rule.UserID] = true
for k := range rule.Labels {
if !labelNamePattern.MatchString(k) {
return fmt.Errorf("访问策略包含无效标签名: %s", k)
}
}
}
return nil
}
func DeleteLogSource(id uint) error {
db := repository.GetDB()
db.Lock()
defer db.Unlock()
for i := range db.LogSources {
if db.LogSources[i].ID == id {
db.LogSources = append(db.LogSources[:i], db.LogSources[i+1:]...)
groups := db.LogPolicyGroups[:0]
for _, group := range db.LogPolicyGroups {
if group.SourceID != id {
groups = append(groups, group)
}
}
db.LogPolicyGroups = groups
db.RequestSave()
return nil
}
}
return fmt.Errorf("数据源不存在")
}
// reachableEndpoint 把部署记录里的访问地址转换成「平台侧能真正连上」的地址。
//
// 主机模式创建部署时存的是容器的监听地址(0.0.0.0:3100 / [::]:3100),
// 那是容器内部视角,平台进程在别的机器/集群里,用它必然 connection refused。
// 换成该主机的真实 IP 再返回;主机信息缺失时原样返回(至少不会更差)。
func reachableEndpoint(d *models.LokiDeployment) string {
endpoint := strings.TrimSpace(d.Endpoint)
if endpoint == "" {
return endpoint
}
if !strings.Contains(endpoint, "0.0.0.0") && !strings.Contains(endpoint, "[::]") {
return endpoint
}
hostIP := ""
if d.Host != nil {
hostIP = strings.TrimSpace(d.Host.HostIP)
}
if hostIP == "" {
// 没带 Host 关联时回库查一次,避免调用方漏预加载
db := repository.GetDB()
if db != nil && d.HostID != nil {
db.RLock()
for i := range db.Hosts {
if db.Hosts[i].ID == *d.HostID {
hostIP = strings.TrimSpace(db.Hosts[i].HostIP)
break
}
}
db.RUnlock()
}
}
if hostIP == "" {
return endpoint
}
endpoint = strings.Replace(endpoint, "0.0.0.0", hostIP, 1)
endpoint = strings.Replace(endpoint, "[::]", hostIP, 1)
return endpoint
}
// EnsureLogSourceForDeployment 部署成功后自动注册查询数据源(host→url;k8s→cluster-svc)。
// 已存在同名数据源时:平台托管的自动刷新连接信息,用户手填的不覆盖。
func EnsureLogSourceForDeployment(d *models.LokiDeployment) {
if d == nil || d.Endpoint == "" {
return
}
svc := "loki"
port := 3100
target := &models.LogSource{
Name: d.Name, TargetType: "cluster-svc",
ClusterID: d.ClusterID, Namespace: d.Namespace, Service: svc, Port: port,
AuthType: "none", Managed: true,
}
if d.Mode == "host-binary" || d.Mode == "host-docker" {
target.TargetType = "url"
// 主机模式部署的 endpoint 记的是**容器监听地址**(0.0.0.0:3100),
// 直接拿来当数据源 URL 会得到 dial tcp 0.0.0.0:3100: connection refused ——
// 平台进程不在那台主机上,0.0.0.0 只在本机有意义。
// 这里换成主机真实 IP,平台才能连上。
target.URL = reachableEndpoint(d)
target.ClusterID = nil
}
db := repository.GetDB()
db.Lock()
for i := range db.LogSources {
if db.LogSources[i].Name != target.Name {
continue
}
if !db.LogSources[i].Managed {
db.Unlock()
return
}
db.LogSources[i].TargetType = target.TargetType
db.LogSources[i].ClusterID = target.ClusterID
db.LogSources[i].Namespace = target.Namespace
db.LogSources[i].Service = target.Service
db.LogSources[i].Port = target.Port
db.LogSources[i].URL = target.URL
db.LogSources[i].UpdatedAt = time.Now()
db.Unlock()
db.RequestSave()
return
}
db.Unlock()
_, _ = CreateLogSource(target)
}
// EnsureLogSourcesForRunningDeployments 补齐运行中部署的数据源,避免查询页默认没有数据源。
func EnsureLogSourcesForRunningDeployments() {
db := repository.GetDB()
db.RLock()
running := make([]models.LokiDeployment, 0, len(db.Deployments))
for i := range db.Deployments {
d := db.Deployments[i]
if d.Status == "running" && d.Endpoint != "" {
running = append(running, d)
}
}
db.RUnlock()
for i := range running {
EnsureLogSourceForDeployment(&running[i])
}
}