容器部署 VictoriaMetrics

VictoriaMetrics（简称 VM）是兼容 Prometheus 生态的时序数据库，资源占用低于 Prometheus，支持单机版与集群版。单机版可直接替换 Prometheus 存储；集群版通过 vmstorage / vminsert / vmselect 拆分读写与存储，水平扩展能力更强。

1. VictoriaMetrics 单节点

单机版（victoria-metrics）集采集写入、存储、查询于一体，适合中小规模监控、作为 Prometheus remote_write 远端存储。

1.1 配置环境变量

1
2
3
4
5
6
7
export CONTAINER_CLI=nerdctl
export VM_TAG=v1.147.0
export VM_IMAGE=victoriametrics/victoria-metrics:$VM_TAG
export VM_INSTANCE_NAME=victoriametrics
export VM_DATA=/data/ops/victoriametrics/$VM_INSTANCE_NAME
export VM_HTTP_PORT=8428
export VM_RETENTION=15d

1.2 启动服务

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
mkdir -p $VM_DATA/data

$CONTAINER_CLI run -d \
  --name $VM_INSTANCE_NAME \
  --restart always \
  --security-opt apparmor=unconfined \
  --security-opt seccomp=unconfined \
  --network host \
  --ulimit nofile=1048576:1048576 \
  --memory-swappiness=0 \
  -v $VM_DATA/data:/victoria-metrics-data \
  $VM_IMAGE \
  -storageDataPath=/victoria-metrics-data \
  -retentionPeriod=$VM_RETENTION \
  -httpListenAddr=:$VM_HTTP_PORT

1.3 验证

1
2
3
4
5
6
7
8
# 健康检查
curl -s http://127.0.0.1:$VM_HTTP_PORT/health

# 查看自身指标
curl -s http://127.0.0.1:$VM_HTTP_PORT/metrics | head

# PromQL 查询（兼容 Prometheus API）
curl -s 'http://127.0.0.1:8428/api/v1/query?query=up'

写入测试（Prometheus remote_write 格式）：

1
2
3
curl -s -X POST http://127.0.0.1:$VM_HTTP_PORT/api/v1/import/prometheus -d '
vm_test_metric{env="dev"} 1
'

1.4 打印交付结果

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
cat <<EOF
VictoriaMetrics (single):
  Instance Name:  $VM_INSTANCE_NAME
  IP:             $(hostname -I | awk '{print $1}')
  HTTP Port:      $VM_HTTP_PORT
  Retention:      $VM_RETENTION
  Write URL:      http://$(hostname -I | awk '{print $1}'):$VM_HTTP_PORT/api/v1/write
  Query URL:      http://$(hostname -I | awk '{print $1}'):$VM_HTTP_PORT/api/v1/query
  Metrics URL:    http://$(hostname -I | awk '{print $1}'):$VM_HTTP_PORT/metrics
EOF

2. VictoriaMetrics 集群

下面以 3 节点 vmstorage + 每台各跑 vminsert / vmselect 为例。这是集群版的最小生产拓扑：vmstorage 存数据，vminsert 负责写入路由，vmselect 负责查询聚合。

节点	IP（示例）	角色
node1	10.0.0.1	vmstorage + vminsert + vmselect
node2	10.0.0.2	vmstorage + vminsert + vmselect
node3	10.0.0.3	vmstorage + vminsert + vmselect

三台机器上都要能互相访问以下端口：

端口	组件	用途
8400	vmstorage	接收 vminsert 写入（TCP）
8401	vmstorage	接收 vmselect 查询（TCP）
8480	vminsert	HTTP 写入 / remote_write
8481	vmselect	HTTP 查询 / Grafana 数据源
8482	vmstorage	HTTP 监控 `/metrics`

2.1 在每个节点配置环境变量

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
export CONTAINER_CLI=nerdctl
export VM_TAG=v1.147.0-cluster
export VM_CLUSTER_NAME=vm_cluster_3n

export VM_STORAGE_IMAGE=victoriametrics/vmstorage:$VM_TAG
export VM_INSERT_IMAGE=victoriametrics/vminsert:$VM_TAG
export VM_SELECT_IMAGE=victoriametrics/vmselect:$VM_TAG

export VM_NODE1_IP=10.0.0.1
export VM_NODE2_IP=10.0.0.2
export VM_NODE3_IP=10.0.0.3

export NODE_ID=1                              # node2 改为 2，node3 改为 3
export VM_NODE_IP=$VM_NODE1_IP                # 本机 IP：node1/2/3 分别填对应变量

export VM_STORAGE_PORT=8400
export VM_SELECT_PORT=8401
export VM_INSERT_HTTP_PORT=8480
export VM_SELECT_HTTP_PORT=8481
export VM_STORAGE_HTTP_PORT=8482
export VM_RETENTION=15d
export VM_REPLICATION_FACTOR=2                # 副本因子，不超过 vmstorage 节点数

export VM_DATA=/data/ops/victoriametrics/$VM_CLUSTER_NAME-node$NODE_ID
export VM_STORAGE_NODES="$VM_NODE1_IP:$VM_STORAGE_PORT,$VM_NODE2_IP:$VM_STORAGE_PORT,$VM_NODE3_IP:$VM_STORAGE_PORT"
export VM_SELECT_NODES="$VM_NODE1_IP:$VM_SELECT_PORT,$VM_NODE2_IP:$VM_SELECT_PORT,$VM_NODE3_IP:$VM_SELECT_PORT"

检查端口是否被占用：

1
lsof -i :$VM_STORAGE_PORT -i :$VM_SELECT_PORT -i :$VM_INSERT_HTTP_PORT -i :$VM_SELECT_HTTP_PORT -i :$VM_STORAGE_HTTP_PORT

2.2 启动 vmstorage

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
mkdir -p $VM_DATA/vmstorage

$CONTAINER_CLI run -d \
  --name vmstorage-$VM_CLUSTER_NAME-$NODE_ID \
  --restart always \
  --security-opt apparmor=unconfined \
  --security-opt seccomp=unconfined \
  --network host \
  --ulimit nofile=1048576:1048576 \
  --memory-swappiness=0 \
  -v $VM_DATA/vmstorage:/storage \
  $VM_STORAGE_IMAGE \
  -storageDataPath=/storage \
  -retentionPeriod=$VM_RETENTION \
  -vminsertAddr=:$VM_STORAGE_PORT \
  -vmselectAddr=:$VM_SELECT_PORT \
  -httpListenAddr=:$VM_STORAGE_HTTP_PORT

2.3 启动 vminsert

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
$CONTAINER_CLI run -d \
  --name vminsert-$VM_CLUSTER_NAME-$NODE_ID \
  --restart always \
  --security-opt apparmor=unconfined \
  --security-opt seccomp=unconfined \
  --network host \
  --ulimit nofile=1048576:1048576 \
  --memory-swappiness=0 \
  $VM_INSERT_IMAGE \
  -storageNode=$VM_STORAGE_NODES \
  -replicationFactor=$VM_REPLICATION_FACTOR \
  -httpListenAddr=:$VM_INSERT_HTTP_PORT

2.4 启动 vmselect

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
$CONTAINER_CLI run -d \
  --name vmselect-$VM_CLUSTER_NAME-$NODE_ID \
  --restart always \
  --security-opt apparmor=unconfined \
  --security-opt seccomp=unconfined \
  --network host \
  --ulimit nofile=1048576:1048576 \
  --memory-swappiness=0 \
  $VM_SELECT_IMAGE \
  -storageNode=$VM_SELECT_NODES \
  -httpListenAddr=:$VM_SELECT_HTTP_PORT

2.5 验证集群

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
# vmstorage 指标
curl -s http://$VM_NODE_IP:$VM_STORAGE_HTTP_PORT/metrics | grep vm_rows | head

# 经 vminsert 写入
curl -s -X POST "http://$VM_NODE_IP:$VM_INSERT_HTTP_PORT/insert/0/prometheus/api/v1/import/prometheus" -d '
vm_cluster_test{node="'"$NODE_ID"'"} 1
'

# 经 vmselect 查询
curl -s "http://$VM_NODE_IP:$VM_SELECT_HTTP_PORT/select/0/prometheus/api/v1/query?query=vm_cluster_test"

查看各 vmstorage 是否都被 vminsert / vmselect 识别：

1
2
curl -s http://$VM_NODE_IP:$VM_INSERT_HTTP_PORT/metrics | grep vm_rpc_vmstorage_is_reachable
curl -s http://$VM_NODE_IP:$VM_SELECT_HTTP_PORT/metrics | grep vm_rpc_vmstorage_is_reachable

vm_rpc_vmstorage_is_reachable 为 1 表示连通正常。

2.6 打印交付结果

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
cat <<EOF
VictoriaMetrics Cluster:
  Cluster Name:       $VM_CLUSTER_NAME
  Node ID:            $NODE_ID
  Node IPs:
    node1:            $VM_NODE1_IP
    node2:            $VM_NODE2_IP
    node3:            $VM_NODE3_IP
  Retention:          $VM_RETENTION
  Replication Factor: $VM_REPLICATION_FACTOR
  Write URLs (任一 vminsert):
    http://$VM_NODE1_IP:$VM_INSERT_HTTP_PORT/insert/0/prometheus/api/v1/write
    http://$VM_NODE2_IP:$VM_INSERT_HTTP_PORT/insert/0/prometheus/api/v1/write
    http://$VM_NODE3_IP:$VM_INSERT_HTTP_PORT/insert/0/prometheus/api/v1/write
  Query URLs (任一 vmselect):
    http://$VM_NODE1_IP:$VM_SELECT_HTTP_PORT/select/0/prometheus/
    http://$VM_NODE2_IP:$VM_SELECT_HTTP_PORT/select/0/prometheus/
    http://$VM_NODE3_IP:$VM_SELECT_HTTP_PORT/select/0/prometheus/
EOF

3. vmagent 采集配置

vmagent 负责抓取 Exporter 指标并 remote_write 到 VM，可替代 Prometheus 采集侧。单机版写 vmsingle；集群版写任意 vminsert。

3.1 配置环境变量

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
export CONTAINER_CLI=nerdctl
export VM_TAG=v1.147.0
export VM_AGENT_IMAGE=victoriametrics/vmagent:$VM_TAG
export VM_AGENT_NAME=vmagent
export VM_AGENT_DATA=/data/ops/victoriametrics/$VM_AGENT_NAME
export VM_AGENT_HTTP_PORT=8429

# 单机版
export VM_REMOTE_WRITE_URL=http://127.0.0.1:8428/api/v1/write

# 集群版（改用 vminsert 地址）
export VM_REMOTE_WRITE_URL=http://10.0.0.1:8480/insert/0/prometheus/api/v1/write

3.2 生成抓取配置

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
mkdir -p $VM_AGENT_DATA

cat > $VM_AGENT_DATA/scrape.yml <<'EOF'
global:
  scrape_interval: 15s

scrape_configs:
  - job_name: node-exporter
    static_configs:
      - targets:
          - 10.0.0.1:9100
EOF

3.3 启动 vmagent

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
$CONTAINER_CLI run -d \
  --name $VM_AGENT_NAME \
  --restart always \
  --security-opt apparmor=unconfined \
  --security-opt seccomp=unconfined \
  --network host \
  --ulimit nofile=1048576:1048576 \
  --memory-swappiness=0 \
  -v $VM_AGENT_DATA/scrape.yml:/etc/vmagent/scrape.yml \
  $VM_AGENT_IMAGE \
  -promscrape.config=/etc/vmagent/scrape.yml \
  -remoteWrite.url=$VM_REMOTE_WRITE_URL \
  -httpListenAddr=:$VM_AGENT_HTTP_PORT

3.4 验证

1
2
curl -s http://127.0.0.1:$VM_AGENT_HTTP_PORT/metrics | grep vmagent_remotewrite_rows_sent_total
curl -s http://127.0.0.1:$VM_AGENT_HTTP_PORT/targets

4. 自监控

VM 各组件均暴露 /metrics，建议由 vmagent 统一抓取并写入集群。

4.1 抓取配置

由 vmagent 统一抓取 VM 组件与 node-exporter，经 remote_write 写入集群。覆盖第 3.2 节配置，三台节点组件均纳入监控：

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
export VM_AGENT_DATA=/data/ops/victoriametrics/vmagent
mkdir -p $VM_AGENT_DATA

cat > $VM_AGENT_DATA/scrape.yml <<'EOF'
global:
  scrape_interval: 15s

scrape_configs:
  - job_name: vmstorage
    static_configs:
      - targets:
          - 10.0.0.1:8482
          - 10.0.0.2:8482
          - 10.0.0.3:8482
        labels:
          cluster: vm_cluster_3n
          component: vmstorage

  - job_name: vminsert
    static_configs:
      - targets:
          - 10.0.0.1:8480
          - 10.0.0.2:8480
          - 10.0.0.3:8480
        labels:
          cluster: vm_cluster_3n
          component: vminsert

  - job_name: vmselect
    static_configs:
      - targets:
          - 10.0.0.1:8481
          - 10.0.0.2:8481
          - 10.0.0.3:8481
        labels:
          cluster: vm_cluster_3n
          component: vmselect

  - job_name: vmagent
    static_configs:
      - targets:
          - 127.0.0.1:8429
        labels:
          cluster: vm_cluster_3n
          component: vmagent

  - job_name: node-exporter
    static_configs:
      - targets:
          - 10.0.0.1:9100
          - 10.0.0.2:9100
          - 10.0.0.3:9100
EOF

单机版将 scrape_configs 改为 vmsingle job（127.0.0.1:8428），去掉 vmstorage / vminsert / vmselect。

修改后重启 vmagent：

1
$CONTAINER_CLI restart $VM_AGENT_NAME

4.2 验证抓取

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
# vmagent 目标状态（State 应为 up）
curl -s http://127.0.0.1:8429/targets | grep -E 'vmstorage|vminsert|vmselect|vmagent'

# 各组件指标抽样
curl -s http://10.0.0.1:8482/metrics | grep -E '^vm_rows|^vm_active_merges' | head
curl -s http://10.0.0.1:8480/metrics | grep vm_rpc_vmstorage_is_reachable
curl -s http://10.0.0.1:8481/metrics | grep vm_cache_requests_total | head

# 经 vmselect 查自监控指标是否入库
curl -s 'http://10.0.0.1:8481/select/0/prometheus/api/v1/query?query=up{component="vmstorage"}'

关注指标：

指标	含义
`vm_rows`	vmstorage 当前时序行数
`vm_rpc_vmstorage_is_reachable`	vminsert/vmselect 到 vmstorage 是否可达（1=正常）
`vmagent_remotewrite_rows_sent_total`	vmagent 成功写入行数
`vm_http_requests_total`	各组件 HTTP 请求量
`process_resident_memory_bytes`	进程内存占用

4.3 导入 Grafana 面板

数据源使用第 5.1 节配置的 Prometheus 类型数据源（指向 vmselect）。

Grafana 左侧 Dashboards → New → Import，输入面板 ID 导入：

面板	ID	适用	说明
VictoriaMetrics - cluster	11176	集群版	vmstorage / vminsert / vmselect RPC、缓存、资源
VictoriaMetrics - vmagent	12683	vmagent	抓取目标、remote_write 速率与错误
VictoriaMetrics - single-node	10229	单机版	vmsingle 写入、查询、存储、资源

导入步骤：

输入 ID（如 11176）→ Load
VictoriaMetrics 数据源选第 5.1 节创建的项（集群填 vmselect URL）
Import

集群版建议同时导入 11176（组件）和 12683（vmagent）；单机版导入 10229。

面板中按 instance 或 component 标签过滤，可分别查看三台节点的 vmstorage 状态。

5. 对接 Grafana / Prometheus

5.1 Grafana 数据源

部署模式	URL
单机版	`http://<vm-ip>:8428`
集群版	`http://<vmselect-ip>:8481/select/0/prometheus/`

类型选 Prometheus，Save & Test 通过即可。

5.2 remote_write 写入

单机版：

1
2
remote_write:
  - url: http://10.0.0.1:8428/api/v1/write

集群版：

1
2
remote_write:
  - url: http://10.0.0.1:8480/insert/0/prometheus/api/v1/write