devopstf / pometheus

Playing around with Prometheus Operator for K8s cluster and apps monitoring

Application workload monitoring using Prometheus Operator (PO)

Yes, you're right, the name of the repo is POmetheus, pee-ou... ... I was trying to highlight the fact that we are using the PO.

Contents

The contents of this repo are composed by a series of K8s manifests that are intended to show how to leverage Prometheus Operator. Apply these YAML files, that have been taken from here to your K8s/OKD/OCP cluster in order to:

• Deploy Prometheus Operator (PO), including RBAC required objects, i.e. clusterRole, clusterRoleBinding, and serviceAccount.
• Deploy the sample app to be scrapped, and create the service.
• Create the serviceMonitor object associated to such a service.
• Create RBAC required objects for Prometheus pods.
• Create a Prometheus object targeting the serviceMonitor previously created via the serviceMonitorSelector and matching labels..
• Create a nodePort service for Prometheus instance, and expose it.
• Create a different serviceMonitor object for a different application in a different namespace.
• Edit the Prometheus object for including the new serviceMonitor through the serviceMonitorNamespaceSelector.

Environment

This demo sample has been tested using a single-node OKD v3.11 cluster deployed on an AWS-provided t2.large instance. I've been using the default namespace, and the system:admin privileged user. Prometheus Operator v0.27.0

Step-by-Step Quickstart Guide

1. Deploy the PO itself:

kubectl apply -f pometheus-sample-deployment.yml

2. Deploy example-app and its associated service:

kubectl apply -f pometheus-sample-app-dp.yml && \
kubectl apply -f pometheus-sample-app-svc.yml

3. Create the ServiceMonitor object:

kubectl apply -f pometheus-sample-service-monitor.yml

4. Create RBAC related objects for Prometheus pods:

kubectl apply -f pometheus-sample-sa.yml && \
kubectl apply -f pometheus-sample-cr.yml && \
kubectl apply -f pometheus-sample-crb.yml

5. Create the Prometheus instance and its associated service:

kubectl apply -f pometheus-sample-prom.yml && \
kubectl apply -f pometheus-sample-prom-svc.yml

6. Expose the Prometheus Service:

oc expose service/prometheus

At this point, you should be able to head to your Prometheus instance's URL for checking that the right targets are getting scrapped. Now you can add another service to be monitored from a different namespace/project. The only thing you need is to create a new serviceMonitor object within that namespace/project, and targeting it into the Prometheus object specs using the serviceMonitorNamespaceSelector.

In my example I'm creating such a serviceMonitor object by applying the pometheus-sample-sm-demoapp.yml manifest that is referring to a demoapplicationnamespace, and a service, labeled as restservice that is associated to this Springboot demo application from Consol Labs.

You can do that editing the Prometheus object:

oc edit prometheus/prometheus

If you go back to your Prometheus UI, you can see how the operator has been doing its magic under the hood, and the new targets, and the associated config file elements appear.

References

• Getting started with Prometheus Operator
• Design document for Prometheus Operator
• Consol Labs Springboot demo application with Prometheus metrics endpoint
• ConSol Open Monitoring Distribution (OMD) Labs Edition