Kubernetes集群Prometheus监控JVM

点击关注公众号

分享到“朋友圈”，每天上午 08:00，干货推送！

1、配置文件config.yaml，默认为搜集所有，可自定义搜集指标

   ---

hostPort: localhost:38080 #定义exporter暴露的端口，自定义

username:

password:

rules:

- pattern: ".*"

2、配置Dockerfile模板，需要将config.yaml及编译产物移动到Dockerfile同级目录下

FROM   harbor.local.com/public/tincere:jdk8-msyh

VOLUME ["/opt"]

WORKDIR /opt

ADD ./target/*.jar     /opt/myservice.jar

ADD config.yaml      /opt/config.yaml

ADD jmx_prometheus_agent.jar /opt/jmx_prometheus_agent.jar

ENTRYPOINT ["java","-Xms512M","-Xmx512M","-Xmn384M","-Xss1M","-XX:MetaspaceSize=256M","-XX:MaxMetaspaceSize=256M","-XX:+UseParNewGC","-XX:+UseConcMarkSweepGC","-XX:CMSInitiatingOccupancyFraction=92","-XX:+UseCMSCompactAtFullCollection","-XX:CMSFullGCsBeforeCompaction=0","-XX:+CMSParallelInitialMarkEnabled","-XX:+CMSScavengeBeforeRemark","-XX:+HeapDumpOnOutOfMemoryError","-XX:HeapDumpPath=/opt","-XX:+PrintGCDetails","-XX:+PrintGCDateStamps","-Xloggc:gc.log","-Dspring.profiles.active=dev",”-javaagent:/opt/jmx_prometheus_agent.jar=38080:config.yaml”,"-jar","/opt/myservice.jar"]

3、服务启动成功后，可以进入Pod内，通过curl获取到jvm metrics  

4、创建服务的Deployment及service, 在服务对应的service将38080端口命名jvm-port

   - name: jvm-port

port: 38080

targetPort: 38080

5、创建serviceMonitor与Prometheus同处一个命名空间，

apiVersion: monitoring.coreos.com/v1

kind: ServiceMonitor

metadata:

  labels:

    k8s-app: jvm-test 

  name: jvm-test

  namespace: kubesphere-monitoring-system

spec:

  endpoints:

  - path: /metrics

    port: jvm-port  

    scheme: http

  namespaceSelector:  

    any: true

  selector: 

    matchLabels:

      app: jvm-test

6、修改kubesphere-monitoring-system空间下名为k8s的Prometheus资源，在serviceMonitorSelector: 字段下添加上步骤4中的serviceMonitor的labels

7、Prometheus会自动发现该serviceMonitor并reload配置文件，可以将Prometheus-k8s通过port-forward映射出来查看

kubectl -n kubesphere-monitoring-system port-forward --address=0.0.0.0 prometheus-k8s-0 39090:9090

当出现上述图片时，说明监控配置成功

8、配置gafana，在grafana官网寻找合适的jvm监控模板，导入，配置jvm监控，效果如图所示