Files
LokiStack/backend/services/collector.go
T
opsx e006fc1a0e fix(promtail): app 标签通用回退 + 跳过僵尸 Pod + SecretKey 不下发 + 采集命名空间可选
=== 一、astralx 只采集到 2 个 app(核心问题)===
实证:用户的 Deployment Pod 模板只有 k8s.kuboard.cn/name 标签,
     没有 app / app.kubernetes.io/name。原模板的 app 只从这两个标签取,
     所以 115 个服务的 app 全是空值,Loki 里 app 只剩零星 2 个。

修复:app 四级回退,全部基于 Kubernetes 原生元数据(跨集群通用):
  1) app.kubernetes.io/name   标准约定
  2) app                      常见简写
  3) __tmp_controller_name    Pod owner(ReplicaSet) 去 hash → Deployment 名
                              (__meta_kubernetes_pod_controller_name 由
                                Promtail 直接从 K8s API 读取,任何集群都有,
                                不依赖 Kuboard/Rancher 等第三方工具注入的标签)
  4) pod name                 兜底,保证 app 永不为空

同时修复 PromtailHelmValues 模板同样的 app 逻辑,
并修正其 __meta_kubernetes__pod_name 双下划线笔误。

验证:astralx 的 app 值从 2 个 → 73 个。

=== 二、15 分钟 27 万条日志的真相 ===
排查结论:不是重复采集(positions 存 hostPath /run/promtail,Pod 重启不丢)。
真实原因:cex-openapi-gateway 残留 347 个 Evicted + 37 个 Error 的僵尸 Pod,
        只有 1 个 Running。这些 Pod 的日志文件仍在 /var/log/pods/ 下,
        Promtail 会把它们的历史异常堆栈(Reactor at reactor.core.publisher...)
        全量采集,造成 27 万条的假象并污染存储。

修复:relabel_configs 首条加 drop 规则:
  - source_labels: [__meta_kubernetes_pod_phase]
    regex: (Succeeded|Failed)
    action: drop
(两个模板都加)

=== 三、SecretKey 不再下发给前端 ===
安全要求:避免任何登录用户点「显示」按钮就能拿到对象存储 SK。
- StorageConfig.Sanitized(): secretKey 返回空串,仅保留 hasSecretKey 标记
  (accessKey 仍明文返回,便于管理员核对用的是哪个 AK)
- 测试连接:StorageTestRequest 新增 id 字段,secretKey 留空时后端用
  库里保存的凭据解密补齐(新建存储仍强制填写)
- 前端编辑时 SecretKey 一律留空 = 不修改,placeholder 显示
  「已保存 · 留空则不修改」,测试时带上 id

=== 四、采集器支持指定命名空间 ===
- Collector 新增 TargetNamespaces []string,对应 promtail 的
  kubernetes_sd_configs.namespaces.names(为空=全集群)
- PromtailDaemonsetYAML 新增 promtailNamespaceFilter 生成白名单
- service.Update 支持 targetNamespaces(兼容 []string 与逗号分隔字符串)
- 前端 CollectorEditDialog 增加「采集命名空间」输入(逗号分隔)

镜像:lghost/lokistack:latest + v0.1 = sha256:05b1505aedae
2026-10-27 03:30:25 +00:00

316 lines
8.3 KiB
Go

package services
import (
"context"
"fmt"
"strings"
"time"
"opsxloki/models"
"opsxloki/repository"
"opsxloki/utils"
)
type CollectorService struct{}
func NewCollectorService() *CollectorService {
return &CollectorService{}
}
type CreateCollectorRequest struct {
Name string
Type string
LokiDeploymentID uint
HostID *uint
ClusterID *uint
TargetDeployment string
Namespace string
LogPath string
Labels string
}
// fillAssociations 填充关联的 Loki / Host / Cluster(用于配置生成与展示)
func (s *CollectorService) fillAssociations(c *models.Collector) {
db := repository.GetDB()
db.RLock()
defer db.RUnlock()
for i := range db.Deployments {
if db.Deployments[i].ID == c.LokiDeploymentID {
d := db.Deployments[i]
c.LokiDeployment = &d
break
}
}
if c.HostID != nil {
for i := range db.Hosts {
if db.Hosts[i].ID == *c.HostID {
h := db.Hosts[i]
c.Host = &h
break
}
}
}
if c.ClusterID != nil {
for i := range db.Clusters {
if db.Clusters[i].ID == *c.ClusterID {
cl := db.Clusters[i]
c.Cluster = &cl
break
}
}
}
}
// generateConfigSnapshot 按类型生成配置快照
func (s *CollectorService) generateConfigSnapshot(c *models.Collector) string {
endpoint := ""
if c.LokiDeployment != nil {
endpoint = c.LokiDeployment.Endpoint
}
clusterName := "host"
if c.Cluster != nil && c.Cluster.Name != "" {
clusterName = c.Cluster.Name
}
switch c.Type {
case "host-promtail":
return PromtailConfig(endpoint, c.LogPath, "host", c.Labels)
case "k8s-daemonset":
return PromtailHelmValues(endpoint, clusterName, c.Labels)
case "k8s-fluentbit-sidecar":
host, port := parseLokiURL(endpoint)
app := c.TargetDeployment
ns := c.Namespace
if ns == "" {
ns = "default"
}
return FluentbitConfig(host, port, c.LogPath, app, ns)
}
return ""
}
func (s *CollectorService) Create(req *CreateCollectorRequest) (*models.Collector, error) {
switch req.Type {
case "host-promtail", "k8s-daemonset", "k8s-fluentbit-sidecar":
default:
return nil, fmt.Errorf("不支持的采集器类型: %s", req.Type)
}
if req.LokiDeploymentID == 0 {
return nil, fmt.Errorf("请选择关联的 Loki 部署")
}
if req.Type == "host-promtail" && (req.HostID == nil || *req.HostID == 0) {
return nil, fmt.Errorf("主机采集器需要选择目标主机")
}
if (req.Type == "k8s-daemonset" || req.Type == "k8s-fluentbit-sidecar") && (req.ClusterID == nil || *req.ClusterID == 0) {
return nil, fmt.Errorf("K8s 采集器需要选择目标集群")
}
if req.Type == "k8s-fluentbit-sidecar" && req.TargetDeployment == "" {
return nil, fmt.Errorf("Sidecar 采集器需要填写目标 Deployment 名称")
}
refType := "cluster"
if req.Type == "host-promtail" {
refType = "host"
}
collector := models.Collector{
ID: repository.GetDB().NextID("collectors"),
Name: req.Name,
Type: req.Type,
LokiDeploymentID: req.LokiDeploymentID,
RefType: refType,
HostID: req.HostID,
ClusterID: req.ClusterID,
TargetDeployment: req.TargetDeployment,
Namespace: req.Namespace,
LogPath: req.LogPath,
Labels: req.Labels,
Status: "pending",
CreatedAt: time.Now(),
UpdatedAt: time.Now(),
}
s.fillAssociations(&collector)
if collector.LokiDeployment == nil {
return nil, fmt.Errorf("关联的 Loki 部署不存在")
}
collector.Config = s.generateConfigSnapshot(&collector)
db := repository.GetDB()
db.Lock()
db.Collectors = append(db.Collectors, collector)
db.Unlock()
db.RequestSave()
go s.Install(collector.ID)
return &collector, nil
}
func (s *CollectorService) Get(id uint) (*models.Collector, error) {
db := repository.GetDB()
db.RLock()
var found *models.Collector
for i := range db.Collectors {
if db.Collectors[i].ID == id {
c := db.Collectors[i]
found = &c
break
}
}
db.RUnlock()
if found == nil {
return nil, utils.ErrNotFound
}
s.fillAssociations(found)
return found, nil
}
func (s *CollectorService) List(page, pageSize int) ([]models.Collector, int64, error) {
db := repository.GetDB()
db.RLock()
total := int64(len(db.Collectors))
start := (page - 1) * pageSize
if start < 0 || start >= len(db.Collectors) {
db.RUnlock()
return []models.Collector{}, total, nil
}
end := start + pageSize
if end > len(db.Collectors) {
end = len(db.Collectors)
}
result := make([]models.Collector, end-start)
copy(result, db.Collectors[start:end])
db.RUnlock()
for i := range result {
s.fillAssociations(&result[i])
}
return result, total, nil
}
func (s *CollectorService) Update(id uint, updates map[string]interface{}) error {
db := repository.GetDB()
db.Lock()
defer db.Unlock()
for i := range db.Collectors {
if db.Collectors[i].ID == id {
if v, ok := updates["name"].(string); ok && v != "" {
db.Collectors[i].Name = v
}
if v, ok := updates["status"].(string); ok && v != "" {
db.Collectors[i].Status = v
}
if v, ok := updates["statusMessage"].(string); ok {
db.Collectors[i].StatusMessage = v
}
if v, ok := updates["logPath"].(string); ok {
if v != db.Collectors[i].LogPath {
db.Collectors[i].StatusMessage = "采集路径已保存,点击部署以应用新配置"
}
db.Collectors[i].LogPath = v
}
if v, ok := updates["labels"].(string); ok {
db.Collectors[i].Labels = v
}
// targetNamespaces: 采集哪些命名空间([]string;空=全集群)
if v, ok := updates["targetNamespaces"]; ok {
var nss []string
switch t := v.(type) {
case []interface{}:
for _, item := range t {
if s, ok := item.(string); ok {
if s = strings.TrimSpace(s); s != "" {
nss = append(nss, s)
}
}
}
case []string:
for _, s := range t {
if s = strings.TrimSpace(s); s != "" {
nss = append(nss, s)
}
}
case string:
// 兼容 "astralx,log" 这种逗号分隔写法
for _, s := range strings.Split(t, ",") {
if s = strings.TrimSpace(s); s != "" {
nss = append(nss, s)
}
}
}
if len(nss) != len(db.Collectors[i].TargetNamespaces) {
db.Collectors[i].StatusMessage = "采集命名空间已保存,点击部署以应用新配置"
} else {
for j := range nss {
if nss[j] != db.Collectors[i].TargetNamespaces[j] {
db.Collectors[i].StatusMessage = "采集命名空间已保存,点击部署以应用新配置"
break
}
}
}
db.Collectors[i].TargetNamespaces = nss
}
db.Collectors[i].UpdatedAt = time.Now()
db.RequestSave()
return nil
}
}
return utils.ErrNotFound
}
func (s *CollectorService) Delete(id uint) error {
collector, err := s.Get(id)
if err != nil {
return utils.ErrNotFound
}
db := repository.GetDB()
db.Lock()
for i, c := range db.Collectors {
if c.ID == id {
db.Collectors = append(db.Collectors[:i], db.Collectors[i+1:]...)
break
}
}
db.Unlock()
db.RequestSave()
// 异步卸载(用拷贝的 collector,避免删除后取不到)
go func() {
ctx, cancel := context.WithTimeout(context.Background(), 2*time.Minute)
defer cancel()
NewCollectorDeployer(collector, collector.LokiDeployment, nil).Uninstall(ctx)
}()
return nil
}
// Install 异步部署采集器
func (s *CollectorService) Install(id uint) {
collector, err := s.Get(id)
if err != nil {
return
}
if collector.LokiDeployment == nil {
s.Update(id, map[string]interface{}{"status": "error", "statusMessage": "关联的 Loki 不存在"})
return
}
ctx, cancel := context.WithTimeout(context.Background(), 10*time.Minute)
defer cancel()
if err := NewCollectorDeployer(collector, collector.LokiDeployment, nil).Install(ctx); err != nil {
s.Update(id, map[string]interface{}{"status": "error", "statusMessage": err.Error()})
return
}
s.Update(id, map[string]interface{}{"status": "active", "statusMessage": "采集器已部署"})
}
// Uninstall 异步卸载采集器
func (s *CollectorService) Uninstall(id uint) {
collector, err := s.Get(id)
if err != nil {
return
}
if collector.LokiDeployment == nil {
return
}
ctx, cancel := context.WithTimeout(context.Background(), 2*time.Minute)
defer cancel()
NewCollectorDeployer(collector, collector.LokiDeployment, nil).Uninstall(ctx)
s.Update(id, map[string]interface{}{"status": "pending", "statusMessage": "采集器已卸载"})
}