In this example we will see how we can expose a the metrics created by the random generator REST service in a Prometheus conformant format.

By intention the metrics created by the the random-generator service is written is an own format which can’t be used by Prometheus directly. Also the metrics are not exposed via a network port but is written into the filesystem. We are using an Adapter sidecar container to convert and expose data written in that custom proprietary format over HTTP so that it can be scraped by Prometheus.

Let’s see how this works.

We are using a bare Pod for simplicities sake (in real world scenarios this would be of course e.g. a Deployment).

Please have a look into the pod.yml descriptor which

• Create two containers: One is our random-generator the other is the Adapter, k8spatterns/random-generator-exporter. This is a simple Perl script (yeah, Perl ;) which does the transformation. You find the Dockerfile and script for this image in the link::image/[image] directory.

• Create a volume which is shared between the main application container and the adapter. The main container writes the metrics file into this shared directory, and the Prometheus Adapter picks it up when queried over HTTP

The Pod can be created with

For easy access of random generator, let’s create a Service:

This Service uses a nodePort to expose it on every node of the cluster. For Minikube this works nicely. However, if your cluster nodes are not exposed, you might either use our kubectl port-forward hack used in other examples or use a full ingress or loadbalancer exposed route as described in the Service Discovery pattern.

Let’s assume that you are using Minikube for now, so that we can access the Then let’s access our Service with

When we now jump into the Pod we can examine the generated metrics file

The metrics extracted from that data can now be accessed in a Prometheus conformant way: