This tutorial demonstrates how to enable monitoring metrics for RadonDB MySQL clusters.

Overview

The text-based format for exposing metrics adopted by Prometheus has been a de facto standard in cloud-native monitoring.

The RadonDB MySQL monitoring engine is based on Prometheus MySQLd Exporter. It scrapes RadonDB MySQL metrics with mysqld-exporter and visualizes the metrics by third-party platforms.

Prerequisites

• You need to prepare a Kubernetes or KubeSphere cluster.
• You should use RadonDB MySQL Operator 2.1.0 or a later version.

Procedure

Step 1 Configure serviceMonitor.

The serviceMonitor field defines the automatic monitoring engine of RadonDB MySQL Operator. After being enabled, the Operator is automatically connected to mysqld_exporter and Prometheus.

You are allowed to configure the serviceMonitor field in the charts/mysql-operator/values.yaml configuration file.

The serviceMonitor field contains the following subfields:

serviceMonitor:
  enabled: true
  ## Additional labels for the serviceMonitor. Useful if you have multiple prometheus operators running to select only specific ServiceMonitors
  # additionalLabels:
  #   prometheus: prom-internal
  interval: 10s
  scrapeTimeout: 3s
  # jobLabel:
  # targetLabels:
  # podTargetLabels:
  namespaceSelector:
    any: true
  selector:
    matchLabels:
      app.kubernetes.io/managed-by: mysql.radondb.com
      app.kubernetes.io/name: mysql

• When a new Operator is deployed, serviceMonitor.enabled is set to true by default, which means the serviceMonitor is enabled.
• If you have an older operator version installed, you need to deploy a supported version of the Operator.

Step 2 Configure metricsOpts.

The metricsOpts field defines the RadonDB MySQL cluster monitoring in the mysqlclusters.mysql.radondb.com CRD. You can enable the monitoring service by configuring the field in the mysql_v1alpha1_mysqlcluster.yaml configuration file.

The metricsOpts field contains the following subfields:

metricsOpts:
    enabled: false  
    image: prom/mysqld-exporter:v0.12.1

    resources:
      requests:
        cpu: 10m
        memory: 32Mi
      limits:
        cpu: 100m
        memory: 128Mi

etricsOpts.enabled is set to false by default. You can set it to true to enable the monitoring function.

• To enable the cluster monitoring function, set metricsOpts.enabled to true.
• To configure the resource quota for monitoring containers, set the resources field.

Apply the configuration and the following information is displayed.

$ kubectl apply -f config/sample/mysql_v1alpha1_mysqlcluster.yaml
cluster.mysql.radondb.com/sample created/configured

View monitoring service

On client

You can view the cluster monitoring service and information of serviceMonitor as follows.

$ kubectl get service,servicemonitor

$ kubectl describe servicemonitor <serviceName>

Expected output

$ kubectl get service,servicemonitor
NAME                             TYPE        CLUSTER-IP      EXTERNAL-IP   PORT(S)    AGE
service/mysql-operator-metrics   ClusterIP   10.96.242.205   <none>        8443/TCP   3h25m
service/sample-follower          ClusterIP   10.96.2.234     <none>        3306/TCP   21h
service/sample-leader            ClusterIP   10.96.30.238    <none>        3306/TCP   21h
service/sample-metrics           ClusterIP   10.96.7.222     <none>        9104/TCP   3h24m
service/sample-mysql             ClusterIP   None            <none>        3306/TCP   21h

NAME                                                              AGE
servicemonitor.monitoring.coreos.com/demo-mysql-operator          3h25m

$ kubectl describe servicemonitor demo-mysql-operator 
Name:         test-radondb-mysql-metrics
Namespace:    default
Labels:       app=test-radondb-mysql
              app.kubernetes.io/managed-by=Helm
              app.kubernetes.io/vendor=kubesphere
              chart=radondb-mysql-1.0.0
              heritage=Helm
              release=test
Annotations:  kubesphere.io/creator: admin
API Version:  monitoring.coreos.com/v1
Kind:         ServiceMonitor
......
Spec:
  Endpoints:
    Interval:        1m
    Path:            /metrics
    Port:            metrics
    Scheme:          http
    Scrape Timeout:  10s
......

On KubeSphere

After the monitoring is enabled, you can view the monitoring service status for RadonDB MySQL Operator and clusters deployed in KubeSphere workspace.

• In the project space, go to Application Workloads > Services, and click <clusterName>-metrics to view the monitoring service details.

• On the Pods page under Application Workloads in the project space, click a container name to view the metrics status of the container.

View application monitoring metrics

Customize monitoring on KubeSphere

Note

RadonDB MySQL Operator and clusters need to be deployed on KubeSphere.

The KubeSphere monitoring engine is based on Prometheus and Prometheus Operator. KubeSphere custom monitoring supports visualizing RadonDB MySQL metric data.

Step 1 In the same project, go to Monitoring & Alerting > Custom Monitoring, and click Create.

Step 2 In the displayed dialog box, set a name for the dashboard (for example, mysql-overview) and select the MySQL template. Click Next to continue.

Step 3 Click Save Template in the upper-right corner. A newly-created dashboard is displayed on the Custom Monitoring Dashboards page.

Step 4 Wait about ten minutes to view the monitoring metrics.

For more information, see KubeSphere Custom Application Monitoring and Visualization.

By Prometheus and Grafana platforms

Grafana is an open-source interactive data-visualization platform. You can view the monitoring information by using Prometheus and Grafana platforms. The following is the process of displaying monitoring metrics by Prometheus and Grafana.

• Obtain the monitoring metric data of RadonDB MySQL services by MySQL server exporter.
• Obtain the monitoring metric data of RadonDB MySQL servers by node exporter.
• Transfer monitoring metric data to Prometheus and configure the data source to display monitoring charts and warnings on Grafana.

For more instructions on Grafana monitoring visualization, see Grafana dashboards.