Service Monitor Operator

Overview

The Service Monitor Operator will create, destroy and update service monitor resources for you when you create, update or destroy a service.

Why

The prometheus operator leverages Service Monitors Custom Resrouces to define what services to monitor and details around how to scrape them. The goal of this project is to try and solve the most common service monitor configurations by automatically creating them for you based on annotations on your service.

Use

The Service Monitor Operator will look at the following annotations on a service.

• prometheus.io/probe: - true or false
• prometheus.io/port: - TCP port listening for scrape requests
• prometheus.io/path: - path to metrics endpoint
• prometheus.io/scheme: - http or https
• prometheus.io/interval:- scrape interval
• prometheus.io/bearertoken: path to bearer token file

These are all optional except for prometheus.io/probe. If any values are omitted, the following are the default values that will be used

• port - 8080
• path - /metrics
• scheme - http
• scrapeInterval - 30s
• bearerTokenFile - /var/run/secrets/kubernetes.io/serviceaccount/token

If the prometheus.io/probe: annotation valuel is not set to true it will ignore this service.
The path annotation is the path to the endpoint that will provide metrics. For example in the case of foo.bar.com/metrics the path is /metrics

The Selectors for the servicemonitor will be set to match the Labels on the service and the Namespace that the service is running in.

Unimplmented parts of the ServiceMonitor spec

The following items are not yet implmented

• TargetLabels are not able to be changed
• PodTargetLabels are not able to be changed
• TLSConfig is not supported

Deploy to cluster

kubectl apply -f deploy -n monitoring

Build

• Requires operator-sdk from the operator framework
• To build an image for deployment/testing run the following
  • operator-sdk build <docker image name of your choosing>
• Update deploy/operator.yaml file with your image name
• Deploy to cluster kubectl apply -f deploy -n monitoring

You can also test without building and deploying to a cluster with the following command operator-sdk up local --namespace=<your namespace> --kubeconfig=<path to your kubeconfig file>

You can ommit the kubeconfig flag if you are using the default kubeconfig on your system ~/.kube/config.