Fabric8 Documentation

Metrics

Capturing historical metrics an extremely useful tool to help diagnose your running system.

With fabric8 we recommend using:

  • Prometheus as the back end storage of historical metric data as they are both easy to scale, provide options for resilience and are very easy to query.
  • Grafana as the console to view, query and analyze metrics.
  • for Java applications we use Agent Bond to expose JMX metrics for Prometheus

How to use Metrics in fabric8

If you are running Fabric8 with the Fabric8 Console then go to the Apps tab.

  • click the Run... button and select the Prometheus app to and run it.
  • click the Run... button and select the Grafana app to and run it.

Cadvisor runs embedded in the kubelet on each node to ensure the metrics are collected.

Once the above is running, the Fabric8 Console should have a Metrics tab letting you view and search the metrics via Grafana.

How to enable metrics scraping in your applications

When using Prometheus to collect metrics for monitoring your containers you can enable scraping in Java Maven projects in the pom.xml file accordingly. For example to enable scraping and service enable this on port 9779, define the following in the pom.xml file:

<fabric8.metrics.scrape>true</fabric8.metrics.scrape>
<fabric8.metrics.port>9779</fabric8.metrics.port>
<fabric8.metrics.scheme>http</fabric8.metrics.scheme>

There must be at least one service define in your application to ensure the metrics scaping is enabled. If no services has been defined, then a headless service must be configured to ensure the metrics scraping works. Add the following two lines to the pom.xml file:

<fabric8.service.name>${project.artifactId}</fabric8.service.name>
<fabric8.service.headless>true</fabric8.service.headless>

This headless service is used to collect application level metrics for ingestion into Prometheus.