1
# Grafana 生产配置
2
apiVersion: 1
3
providers:
4
  - name: 'default'
5
    orgId: 1
6
    folder: ''
7
    type: file
8
    disableDeletion: false
9
    editable: true
10
    options:
11
      path: /var/lib/grafana/dashboards
12
      foldersFromFilesStructure: true
13

14
datasources:
15
  - name: Mimir
16
    type: prometheus
17
    uid: mimir
18
    url: http://mimir:9009/prometheus
19
    jsonData:
20
      httpHeaderName1: X-Scope-OrgID
21
    secureJsonData:
22
      httpHeaderValue1: ${TENANT_ID}
23
    exemplarTraceIdDestinations:
24
      - name: trace_id
25
        datasourceUid: tempo
26

27
  - name: Tempo
28
    type: tempo
29
    uid: tempo
30
    url: http://tempo:3200
31
    jsonData:
32
      tracesToMetrics:
33
        datasourceUid: mimir
34
      tracesToLogs:
35
        datasourceUid: loki
36
        filterByTraceID: true
37

38
  - name: Loki
39
    type: loki
40
    uid: loki
41
    url: http://loki:3100
42
    jsonData:
43
      derivedFields:
44
        - name: TraceID
45
          matcherRegex: '"trace_id":"(\w+)"'
46
          datasourceUid: tempo

1
# Mimir 多租户配置
2
multitenancy_enabled: true
3

4
limits:
5
  # 每租户限制
6
  max_global_series_per_user: 5000000
7
  ingestion_rate: 250000
8
  ingestion_burst_size: 500000
9

10
  # 按租户覆盖
11
  per_tenant_override_config: /etc/mimir/overrides.yaml

1
# overrides.yaml — 租户覆盖
2
overrides:
3
  tenant-payment:
4
    max_global_series_per_user: 10000000
5
    ingestion_rate: 500000
6
  tenant-experimental:
7
    max_global_series_per_user: 100000
8
    ingestion_rate: 10000

1
# Loki 多租户
2
auth_enabled: true
3

4
limits_config:
5
  max_streams_per_user: 10000
6
  max_global_streams_per_user: 50000
7
  ingestion_rate_mb: 20
8
  ingestion_burst_size_mb: 30

1
# Grafana RBAC 配置
2
# 组织角色
3
roles:
4
  - name: viewer
5
    permissions:
6
      - dashboards:read
7
      - datasources:query
8
      - explore:use
9

10
  - name: editor
11
    permissions:
12
      - dashboards:read
13
      - dashboards:write
14
      - datasources:query
15
      - explore:use
16
      - alerts:read
17

18
  - name: admin
19
    permissions:
20
      - dashboards:read
21
      - dashboards:write
22
      - datasources:query
23
      - datasources:write
24
      - explore:use
25
      - alerts:read
26
      - alerts:write
27
      - users:manage
28
      - org:manage
29

30
# 团队级权限
31
teams:
32
  - name: order-team
33
    members: ["alice", "bob", "charlie"]
34
    role: editor
35
    folders:
36
      - order-service  # 只能编辑 order-service 文件夹下的面板
37

38
  - name: sre-team
39
    members: ["dave", "eve"]
40
    role: admin
41
    folders:
42
      - "*"  # 所有文件夹

1
# Tempo 保留策略
2
compactor:
3
  compaction:
4
    block_retention: 720h  # 30 天
5
    compaction_window: 1h
6
    max_block_bytes: 100MB
7
    block_ranges:
8
      - 2h
9
      - 12h
10
      - 24h

1
# Mimir 生命周期
2
blocks_storage:
3
  bucket_store:
4
    # 对象存储配置
5
    type: s3
6
    s3:
7
      endpoint: s3.amazonaws.com
8
      bucket_name: mimir-blocks
9
  tsdb:
10
    # 本地数据保留（热数据）
11
    retention_period: 7d
12

13
compactor:
14
  # 降采样和压缩
15
  compaction_interval: 30m
16
  downsampling:
17
    - resolution: 5m
18
      retention: 30d
19
    - resolution: 1h
20
      retention: 365d

1
# Loki 生命周期
2
schema_config:
3
  configs:
4
    - from: 2024-01-01
5
      store: tsdb
6
      object_store: s3
7
      schema: v13
8

9
compactor:
10
  working_directory: /loki/compactor
11
  compaction_interval: 10m
12
  retention_enabled: true
13
  retention_delete_delay: 2h
14
  delete_request_cancel_period: 24h
15

16
limits_config:
17
  # 日志保留策略
18
  retention_period: 14d  # 14 天后删除
19
  # 按流保留
20
  max_query_length: 720h

1
# Tempo 生命周期
2
compactor:
3
  compaction:
4
    block_retention: 720h  # 30 天
5
    compaction_window: 1h
6
    max_block_bytes: 100MB
7

8
storage:
9
  trace:
10
    backend: s3
11
    s3:
12
      bucket: tempo-traces
13
    # 本地缓存
14
    cache: 1GB

1
# 每租户配额
2
limits:
3
  tenant-default:
4
    max_global_series_per_user: 5000000
5
    ingestion_rate: 250000
6
    max_query_length: 720h
7
  tenant-payment:
8
    max_global_series_per_user: 10000000
9
    ingestion_rate: 500000

1
# 按团队的指标序列数
2
count by (team) ({__name__=~".+"})
3

4
# 按团队的日志写入速率
5
sum by (team) (rate(loki_distributor_lines_received_total[5m]))
6

7
# 按团队的追踪写入速率
8
sum by (team) (rate(tempo_distributor_spans_received_total[5m]))
9

10
# 估算成本（按序列数）
11
count by (team) ({__name__=~".+"}) * 0.0001  # $0.1/千序列/月

1
# 配置 Mimir 多租户
2
# 在 HTTP 头部中传递 X-Scope-OrgID
3
curl -H "X-Scope-OrgID: team-order" http://localhost:9009/prometheus/api/v1/query?query=up

1
# 创建团队
2
curl -X POST http://localhost:3000/api/teams \
3
  -H "Authorization: Bearer $GRAFANA_API_KEY" \
4
  -H "Content-Type: application/json" \
5
  -d '{"name": "order-team", "email": "order-team@company.com"}'
6

7
# 添加成员
8
curl -X POST http://localhost:3000/api/teams/1/members \
9
  -H "Authorization: Bearer $GRAFANA_API_KEY" \
10
  -H "Content-Type: application/json" \
11
  -d '{"userId": 2}'
12

13
# 设置文件夹权限
14
curl -X POST http://localhost:3000/api/folders/order-service/permissions \
15
  -H "Authorization: Bearer $GRAFANA_API_KEY" \
16
  -H "Content-Type: application/json" \
17
  -d '{"items": [{"role": "Editor", "permission": 2}]}'