Azure VM (44) 使用Azure Monitor Metric Exporter

　　《Windows Azure Platform 系列文章目录》

　　一.用户现状及需求

　　1.客户团队使用Prometheus Cloud Watch Exporter，把AWS监控指标，与Prometheus整合：

　　https://github.com/prometheus/cloudwatch_exporter

　　客户团队希望微软云Azure也提供类似的Exporter功能，能把Azure的监控指标(虚拟机、Redis PaaS等，MySQL PaaS数据等)，与Prometheus整合

　　二.说明

　　微软目前没有官方提供的Exporter功能，但是查询到第三方开源的解决方案：
　　https://github.com/webdevops/azure-metrics-exporter

　　三.技术实现

　　该方案通过基于Azure SDK for Go，实现Azure Monitor Metric Exporter功能

　　四.实现关键步骤

　　1.创建和使用Azure订阅，步骤略

　　2.创建Service Principal，并赋权的权限为订阅的Reader。具体步骤略。

　　3.安装Azure虚拟机，我们这里以CentOS 7.9为例，具体步骤略

　　4.设置环境变量

vi ~/.bashrc

　　5.设置Service Principal相关信息

#App ID
export AZURE_CLIENT_ID=XXXXXXXX

#租户ID
export AZURE_TENANT_ID=XXXXXXXX

#App Key
export AZURE_CLIENT_SECRET=XXXXXXXX

　　6.设置环境变量生效

source ~/.bashrc

　　五.安装Prometheus

　　1.我这里使用Prometheus 2.50.1，具体安装步骤略

　　2.下载与运行Azure Monitor Metric Exporter项目，项目文件在：https://github.com/webdevops/azure-metrics-exporter/releases

　　3.我们先下载24.2.0版本：

　　https://github.com/webdevops/azure-metrics-exporter/releases/download/24.2.0/azure-metrics-exporter.linux.amd64

　　4.下载后运行

nohup ./azure-metrics-exporter.linux.amd64 &

　　六.配置Prometheus yml文件for Azure Storage

　　我们编辑prometheus.yml文件，增加下面的内容
　　1. job_name，设置job名称
　　2. 下面的第7行，是我的订阅ID。请PE团队按照实际情况修改
　　3. 下面的第11行，是指标名称。我们这里查询的是BlobCapacity，存储容量大小
　　4. 请注意下图使用的端口号，在第21行，为8080

　　具体指标可以参考文档：https://github.com/webdevops/azure-metrics-exporter

  - job_name: azure-metrics-storageaccount-connections
    scrape_interval: 1m
    metrics_path: /probe/metrics/list
    params:
      name: ["my_own_metric_name"]
      subscription:
      - 166157a8-9ce9-400b-91c7-1d42482b83d6
      resourceType: ["Microsoft.Storage/storageAccounts"]
      metricNamespace: ["Microsoft.Storage/storageAccounts/blobServices"]
      metric:
      - BlobCapacity
      interval: ["PT1H"]
      timespan: ["PT1H"]
      aggregation:
      - average
      - count
      # by blobtype (dimension support)
      # metricFilter: ["BlobType eq '*'"]
      metricTop: ["10"]
    static_configs:
    - targets: ["localhost:8080"]

　　七.配置Prometheus yml文件for Azure MySQL Flexible Server

　　1. 以下是配置Azure MySQL Flexible Server的Prometheus配置文件

　　2. 具体的Metric可以参考：https://learn.microsoft.com/en-us/azure/azure-monitor/reference/supported-metrics/microsoft-dbformysql-flexibleservers-metrics

  - job_name: azure-metrics-databases
    scrape_interval: 1m
    metrics_path: /probe/metrics/list
    params:
      name: ["azure-database"]
      subscription:
      - 166157a8-9ce9-400b-91c7-1d42482b83d6
      #filter: ["resourceType eq 'Microsoft.DBforMySQL/servers'"]
      resourceType: ["Microsoft.DBforMySQL/flexibleServers"]
      #metricNamespace: ["Microsoft.DBforMySQL/flexibleServers"]
      metric:
      - cpu_percent
      - memory_percent
      interval: ["PT1M"]
      timespan: ["2024-08-09T07:00:00Z/2024-08-09T08:00:00Z"]
      aggregation:
      - average
      #- count
      # by blobtype (dimension support)
      # metricFilter: ["BlobType eq '*'"]
      metricTop: ["10"]
    static_configs:
    - targets: ["localhost:8080"]

　　八.Azure Postgre SQL Flexible Server

　　1.以下是配置Azure PGSQL Flexible Server的Prometheus配置文件

　　2.主要监控的指标有两个：CPU利用率和内存利用率

　　3.具体的Metric可以参考：https://learn.microsoft.com/en-us/azure/azure-monitor/reference/supported-metrics/microsoft-dbforpostgresql-flexibleservers-metrics

  - job_name: azure-metrics-pgsql
    scrape_interval: 1m
    metrics_path: /probe/metrics/list
    params:
      name: ["azure-pgsql"]
      subscription:
      - 166157a8-9ce9-400b-91c7-1d42482b83d6
      resourceType: ["Microsoft.DBforPostgreSQL/flexibleServers"]
      metric:
      - cpu_percent
      - memory_percent
      interval: ["PT1M"]
      #P7D表示最近7天
      timespan: ["P7D"]
      aggregation:
      - average
      #- count
      metricTop: ["20"]
    static_configs:
    - targets: ["localhost:8080"]

　　九.根据PGSQL Flexible Server名称等于某个值

　　1. 以下是配置Azure PGSQL Flexible Server的Prometheus配置文件
　　2. 显示类型为：PGSQL Flexible Server
　　3. 筛选PGSQL的Server Name为：等于leipgsql01
　　4. 这里用的是OData的运算符eq，也就是equal，等于某个值

  - job_name: azure-metrics-pgsql
    scrape_interval: 1m
    metrics_path: /probe/metrics/list
    params:
      name: ["azure-pgsql"]
      subscription:
      - 166157a8-9ce9-400b-91c7-1d42482b83d6
      filter: ["resourceType eq 'Microsoft.DBforPostgreSQL/flexibleServers' and name eq 'leipgsql01'"]
      #filter: ["resourceName eq 'leipgsql01'"]
      #resourceType: ["Microsoft.DBforPostgreSQL/flexibleServers"]
      #metricNamespace: ["Microsoft.DBforMySQL/flexibleServers"]
      metric:
      - cpu_percent
      - memory_percent
      interval: ["PT1M"]
      #timespan: ["2024-12-26T03:00:00Z/2024-12-28T08:00:00Z"]
      timespan: ["P7D"]
      aggregation:
      - average
      #- count
      # by blobtype (dimension support)
      # metricFilter: ["BlobType eq '*'"]
      metricTop: ["20"]
    static_configs:
    - targets: ["localhost:8080"]

执行结果，可以看到只显示Server Name为leipgsql01的指标：cpu_percent，memory_percent

# HELP azure_pgsql Azure monitor insight metric
# TYPE azure_pgsql gauge
azure_pgsql{aggregation="average",interval="PT1M",metric="cpu_percent",resourceGroup="sig-rg",resourceID="/subscriptions/166157a8-9ce9-400b-91c7-1d42482b83d6/resourcegroups/sig-rg/providers/microsoft.dbforpostgresql/flexibleservers/leipgsql01",resourceName="leipgsql01",subscriptionID="166157a8-9ce9-400b-91c7-1d42482b83d6",subscriptionName="leizhang-non-prod",tag_owner="",timespan="P7D",unit="Percent"} 10.5
azure_pgsql{aggregation="average",interval="PT1M",metric="memory_percent",resourceGroup="sig-rg",resourceID="/subscriptions/166157a8-9ce9-400b-91c7-1d42482b83d6/resourcegroups/sig-rg/providers/microsoft.dbforpostgresql/flexibleservers/leipgsql01",resourceName="leipgsql01",subscriptionID="166157a8-9ce9-400b-91c7-1d42482b83d6",subscriptionName="leizhang-non-prod",tag_owner="",timespan="P7D",unit="Percent"} 65.5

　　十.根据PGSQL Flexible Server名称不等于某个值

　　1. 以下是配置Azure PGSQL Flexible Server的Prometheus配置文件
　　2. 显示类型为：PGSQL Flexible Server
　　3. 筛选PGSQL的Server Name为：不等于leipgsql01
　　4. 这里用的是OData的运算符ne，也就是not equal，不等于某个值

- job_name: azure-metrics-pgsql
    scrape_interval: 1m
    metrics_path: /probe/metrics/list
    params:
      name: ["azure-pgsql"]
      subscription:
      - 166157a8-9ce9-400b-91c7-1d42482b83d6
      filter: ["resourceType eq 'Microsoft.DBforPostgreSQL/flexibleServers' and name ne 'leipgsql01'"]
      #filter: ["resourceName eq 'leipgsql01'"]
      #resourceType: ["Microsoft.DBforPostgreSQL/flexibleServers"]
      #metricNamespace: ["Microsoft.DBforMySQL/flexibleServers"]
      metric:
      - cpu_percent
      - memory_percent
      interval: ["PT1M"]
      #timespan: ["2024-12-26T03:00:00Z/2024-12-28T08:00:00Z"]
      timespan: ["P7D"]
      aggregation:
      - average
      #- count
      # by blobtype (dimension support)
      # metricFilter: ["BlobType eq '*'"]
      metricTop: ["20"]
    static_configs:
    - targets: ["localhost:8080"]

　　十一.根据PGSQL Flexible Server名称包含某个值

　　1. 以下是配置Azure PGSQL Flexible Server的Prometheus配置文件
　　2. 显示类型为：PGSQL Flexible Server
　　3. 筛选PGSQL的Server Name为：包含lei
　　4. 这里用的是OData的运算符substringof

  - job_name: "prometheus"

    # metrics_path defaults to '/metrics'
    # scheme defaults to 'http'.

    static_configs:
      - targets: ["localhost:9090"]
  - job_name: azure-metrics-pgsql
    scrape_interval: 1m
    metrics_path: /probe/metrics/list
    params:
      name: ["azure-pgsql"]
      subscription:
      - 166157a8-9ce9-400b-91c7-1d42482b83d6
      filter: ["resourceType eq 'Microsoft.DBforPostgreSQL/flexibleServers' and substringof('lei',name)"]
      #filter: ["resourceName eq 'leipgsql01'"]
      #resourceType: ["Microsoft.DBforPostgreSQL/flexibleServers"]
      #metricNamespace: ["Microsoft.DBforMySQL/flexibleServers"]
      metric:
      - cpu_percent
      - memory_percent
      interval: ["PT1M"]
      #timespan: ["2024-12-26T03:00:00Z/2024-12-28T08:00:00Z"]
      timespan: ["P7D"]
      aggregation:
      - average
      #- count
      # by blobtype (dimension support)
      # metricFilter: ["BlobType eq '*'"]
      metricTop: ["20"]
    static_configs:
    - targets: ["localhost:8080"]

　　十二.启动Prometheus

　　1.我们执行命令：

./prometheus --config.file=prometheus.yml

　　2.Prometheus的默认端口为9090

　　3.打开浏览器，查看http://ip:9090，如下图。点击Status, Target

　　4.下图展示的azure-metric-storageaccount-connection，就是我们之前配置的

　　5.我们打开上图的Exporter端口，显示结果为，包含我订阅下所有存储账户名称和存储账户的容量大小

# HELP my_own_metric_name Azure monitor insight metric
# TYPE my_own_metric_name gauge
my_own_metric_name{aggregation="average",interval="PT1H",metric="BlobCapacity",resourceGroup="cdn-rg",resourceID="/subscriptions/166157a8-9ce9-400b-91c7-1d42482b83d6/resourcegroups/cdn-rg/providers/microsoft.storage/storageaccounts/leicdnoriginalstorage",resourceName="leicdnoriginalstorage",subscriptionID="166157a8-9ce9-400b-91c7-1d42482b83d6",subscriptionName="leizhang-non-prod",tag_owner="",timespan="PT1H",unit="Bytes"} 33835
my_own_metric_name{aggregation="average",interval="PT1H",metric="BlobCapacity",resourceGroup="cloud-shell-storage-southeastasia",resourceID="/subscriptions/166157a8-9ce9-400b-91c7-1d42482b83d6/resourcegroups/cloud-shell-storage-southeastasia/providers/microsoft.storage/storageaccounts/cs110032002647d220b",resourceName="cs110032002647d220b",subscriptionID="166157a8-9ce9-400b-91c7-1d42482b83d6",subscriptionName="leizhang-non-prod",tag_owner="",timespan="PT1H",unit="Bytes"} 0
my_own_metric_name{aggregation="average",interval="PT1H",metric="BlobCapacity",resourceGroup="fw-hybrid-test",resourceID="/subscriptions/166157a8-9ce9-400b-91c7-1d42482b83d6/resourcegroups/fw-hybrid-test/providers/microsoft.storage/storageaccounts/niostoragetest01",resourceName="niostoragetest01",subscriptionID="166157a8-9ce9-400b-91c7-1d42482b83d6",subscriptionName="leizhang-non-prod",tag_owner="",timespan="PT1H",unit="Bytes"} 0
my_own_metric_name{aggregation="average",interval="PT1H",metric="BlobCapacity",resourceGroup="lab-rg",resourceID="/subscriptions/166157a8-9ce9-400b-91c7-1d42482b83d6/resourcegroups/lab-rg/providers/microsoft.storage/storageaccounts/leiadls",resourceName="leiadls",subscriptionID="166157a8-9ce9-400b-91c7-1d42482b83d6",subscriptionName="leizhang-non-prod",tag_owner="",timespan="PT1H",unit="Bytes"} 1104
my_own_metric_name{aggregation="average",interval="PT1H",metric="BlobCapacity",resourceGroup="lab-rg",resourceID="/subscriptions/166157a8-9ce9-400b-91c7-1d42482b83d6/resourcegroups/lab-rg/providers/microsoft.storage/storageaccounts/leilabstorage01",resourceName="leilabstorage01",subscriptionID="166157a8-9ce9-400b-91c7-1d42482b83d6",subscriptionName="leizhang-non-prod",tag_owner="",timespan="PT1H",unit="Bytes"} 34272

　　这就是我的环境的5个存储对象，如下图：

　　6.Azure Monitor Metric Exporter还提供调试的功能，我们可以执行：

nohup ./azure-metrics-exporter.linux.amd64 --development.webui &

　　7.Azure metric exporter提供web 界面进行查询。以我的环境为例，打开链接：http://20.52.9.41:8080/query。我们可以在下面进行调试：

posted on 2024-03-15 17:47 Lei Zhang的博客阅读(207) 评论(0) 编辑收藏举报

刷新页面返回顶部

登录后才能查看或发表评论，立即登录或者逛逛博客园首页

相关博文：

· Azure API Management (3) API Management 与虚拟网络VNet集成，并通过Application Gateway进行保护

· Azure Storage (31) Azure File审核下载日志

· Prometheus01 Prometheus基础, 部署与配置, Node Exporter, Pushgateway, PromQL 运算

· Kubernetes-Windows-实用指南-全-

· Azure-Arc-Kubernetes-和服务器教程-全-

阅读排行：
· TypeScript + Deepseek 打造卜卦网站：技术与玄学的结合
· 阿里巴巴 QwQ-32B真的超越了 DeepSeek R-1吗？
· 【译】Visual Studio 中新的强大生产力特性
· 10年+ .NET Coder 心语 ── 封装的思维：从隐藏、稳定开始理解其本质意义
· 【设计模式】告别冗长if-else语句：使用策略模式优化代码结构

历史上的今天：
2018-03-15 Windows Azure Web Site (19) Azure Web App链接到VSTS

Azure Lei Zhang的博客

公告

搜索

常用链接

积分与排名

随笔分类

随笔档案

相册

友情连接

阅读排行榜

评论排行榜

推荐排行榜

最新评论