LokiStack/backend/services/loki_manifest.go

// LokiDeployer — k8s-yaml 模式 manifest 生成(StatefulSet/PVC/ConfigMap/Service)。
package services

import (
	"crypto/sha256"
	"encoding/base64"
	"fmt"

	"opsxloki/pkg/crypto"
)

func (d *LokiDeployer) generateManifest(ns string) string {
	version := d.effectiveVersion()
	image := ResolveImageFor("loki", "grafana/loki", version)
	replicas := d.deployment.Replicas
	if replicas < 1 {
		replicas = 1
	}
	storageSize := d.deployment.StorageSize
	if storageSize == "" {
		storageSize = "20Gi"
	}

	cpu := d.deployment.ResourcesCPU
	mem := d.deployment.ResourcesMem
	if cpu == "" {
		cpu = "1000m"
	}
	if mem == "" {
		mem = "1Gi"
	}

	configData := d.generateLokiConfig()
	secretManifest := ""
	configExpandArg := ""
	storageEnvBlock := ""
	if sc := d.deployment.StorageConfig; sc != nil {
		switch sc.StorageType {
		// 注意：必须包含 cos / obs —— 腾讯云 COS、华为 OBS 同样走 S3 兼容协议，
		// 也需要生成 loki-storage-credentials Secret、注入 LOKI_STORAGE_* 环境变量、
		// 并给 Loki 加 -config.expand-env=true，否则配置里的 ${LOKI_STORAGE_ACCESS_KEY}
		// 无法展开 → AK 为空 → 对象存储 403 InvalidAccessKeyId。
		case "s3", "oss", "cos", "obs", "minio", "azure":
			secretManifest = fmt.Sprintf(`---
apiVersion: v1
kind: Secret
metadata:
  name: loki-storage-credentials
  namespace: %s
type: Opaque
data:
  accessKey: %s
  secretKey: %s`, ns,
				base64.StdEncoding.EncodeToString([]byte(crypto.DecryptField(sc.AccessKey))),
				base64.StdEncoding.EncodeToString([]byte(crypto.DecryptField(sc.SecretKey))))
			configExpandArg = "\n          - -config.expand-env=true"
			storageEnvBlock = `
        env:
        - name: LOKI_STORAGE_ACCESS_KEY
          valueFrom:
            secretKeyRef:
              name: loki-storage-credentials
              key: accessKey
        - name: LOKI_STORAGE_SECRET_KEY
          valueFrom:
            secretKeyRef:
              name: loki-storage-credentials
              key: secretKey`
		}
	}
	// ConfigMap/Secret 更新本身不会重建 Pod。把内容摘要写入 Pod 模板以触发滚动更新。
	configHash := sha256.Sum256([]byte(configData + "\n" + secretManifest))

	manifest := fmt.Sprintf(`apiVersion: v1
kind: Namespace
metadata:
  name: %s
---
apiVersion: v1
kind: ConfigMap
metadata:
  name: loki-config
  namespace: %s
data:
  loki.yaml: |
%s
%s
---
apiVersion: v1
kind: Service
metadata:
  name: loki
  namespace: %s
  labels:
    app: loki
spec:
  type: ClusterIP
  ports:
  - name: http
    port: 3100
    targetPort: 3100
    protocol: TCP
  - name: grpc
    port: 9095
    targetPort: 9095
    protocol: TCP
  selector:
    app: loki
---
apiVersion: v1
kind: PersistentVolumeClaim
metadata:
  name: loki-data
  namespace: %s
spec:
  accessModes:
    - ReadWriteOnce
  resources:
    requests:
      storage: %s
---
apiVersion: apps/v1
kind: StatefulSet
metadata:
  name: loki
  namespace: %s
  labels:
    app: loki
spec:
  serviceName: loki
  replicas: %d
  selector:
    matchLabels:
      app: loki
  template:
    metadata:
      labels:
        app: loki
      annotations:
        checksum/loki-config: "%x"
    spec:
      securityContext:
        runAsUser: 0
        runAsGroup: 0
        fsGroup: 0
      containers:
      - name: loki
        image: %s
        imagePullPolicy: IfNotPresent
        args:
          - -config.file=/etc/loki/loki.yaml%s%s
        ports:
        - name: http
          containerPort: 3100
          protocol: TCP
        - name: grpc
          containerPort: 9095
          protocol: TCP
        resources:
          limits:
            cpu: "%s"
            memory: "%s"
          requests:
            cpu: "500m"
            memory: "512Mi"
        securityContext:
          allowPrivilegeEscalation: false
          readOnlyRootFilesystem: false
          capabilities:
            drop:
            - ALL
        livenessProbe:
          httpGet:
            path: /ready
            port: http
          initialDelaySeconds: 30
          periodSeconds: 10
          timeoutSeconds: 5
          failureThreshold: 6
        readinessProbe:
          httpGet:
            path: /ready
            port: http
          initialDelaySeconds: 10
          periodSeconds: 5
          timeoutSeconds: 3
          failureThreshold: 3
        startupProbe:
          httpGet:
            path: /ready
            port: http
          initialDelaySeconds: 10
          periodSeconds: 10
          timeoutSeconds: 5
          failureThreshold: 30
        volumeMounts:
        - name: config
          mountPath: /etc/loki
          readOnly: true
        - name: data
          mountPath: /var/lib/loki
        - name: wal
          mountPath: /var/lib/loki/wal
      volumes:
      - name: config
        configMap:
          name: loki-config
      - name: data
        persistentVolumeClaim:
          claimName: loki-data
      - name: wal
        emptyDir: {}
`, ns, ns, indentYAML(configData, 4), secretManifest, ns, ns, storageSize, ns, replicas, configHash, image, configExpandArg, storageEnvBlock, cpu, mem)

	// 多副本：追加 Headless Service（clusterIP: None + selector app:loki），
	// 使 memberlist 的 join_members (loki-headless.<ns>.svc.cluster.local) 能解析到所有 Pod IP。
	if replicas >= 2 {
		manifest += fmt.Sprintf(`
---
apiVersion: v1
kind: Service
metadata:
  name: loki-headless
  namespace: %s
  labels:
    app: loki
spec:
  clusterIP: None
  ports:
  - name: http
    port: 3100
    targetPort: 3100
    protocol: TCP
  - name: grpc
    port: 9095
    targetPort: 9095
    protocol: TCP
  - name: memberlist
    port: 7946
    targetPort: 7946
    protocol: TCP
  selector:
    app: loki
`, ns)
	}
	return manifest
}

// ==================== Loki 本地配置生成 (text/template) ====================