Introduction to Custom Application Monitoring

    The KubeSphere monitoring engine is powered by Prometheus and Prometheus Operator. To integrate custom application metrics into KubeSphere, you need to go through the following steps in general.

    • Apply ServiceMonitor CRD to hook up your application with the monitoring target.
    • on the dashboard to view the trend of custom metrics.

    First of all, your application must expose Prometheus-formatted metrics. The Prometheus exposition format is the de-facto format in the realm of cloud-native monitoring. Prometheus uses a text-based exposition format. Depending on your application and use case, there are two ways to expose metrics:

    Direct exposing

    Directly exposing Prometheus metrics from applications is a common way among cloud-native applications. It requires developers to import Prometheus client libraries in their codes and expose metrics at a specific endpoint. Many applications, such as etcd, CoreDNS, and Istio, adopt this method.

    The Prometheus community offers client libraries for most programming languages. Find your language on the Prometheus Client Libraries page. For Go developers, read to learn how to write a Prometheus-compliant application.

    Indirect exposing

    If you don’t want to modify your code or you cannot do so because the application is provided by a third party, you can deploy an exporter which serves as an agent to scrape metric data and translate them into Prometheus format.

    For most third-party applications, such as MySQL, the Prometheus community provides production-ready exporters. Refer to for available exporters. In KubeSphere, it is recommended to enable OpenPitrix and deploy exporters from the App Store. Exporters for MySQL, Elasticsearch, and Redis are all built-in apps in the App Store.

    Please read to learn how to deploy a MySQL exporter and monitor MySQL metrics.

    Writing an exporter is nothing short of instrumenting an application with Prometheus client libraries. The only difference is that exporters need to connect to applications and translate application metrics into Prometheus format.

    The ServiceMonitor CRD is defined by Prometheus Operator. A ServiceMonitor contains information about the metrics endpoints. With ServiceMonitor objects, the KubeSphere monitoring engine knows where and how to scape metrics. For each monitoring target, you apply a ServiceMonitor object to hook your application (or exporters) up to KubeSphere.

    In KubeSphere v3.0.0, you need to pack a ServiceMonitor with your applications (or exporters) into a Helm chart for reuse. In future releases, KubeSphere will provide graphical interfaces for easy operation.

    Please read to learn how to pack a ServiceMonitor with your application.

    Around two minutes, the KubeSphere monitoring engine starts to scape and store metrics. Then you can use PromQL to query metrics and design panels and dashboards.