Prometheus Stack on Kubernetes: Complete Monitoring & Observability Guide

What is Kubernetes Monitoring and Why It Matters?

Kubernetes has emerged as the de facto platform for orchestrating and scaling containerized applications. While it empowers DevOps and SRE teams with automation and agility, Kubernetes also introduces operational complexity. Its highly dynamic, distributed nature means multiple workloads, pods, and services are constantly being created, modified, and terminated, making real-time visibility essential.

Kubernetes Monitoring involves collecting, analyzing, and visualizing metrics across clusters, nodes, and workloads to ensure application health and performance. Without an effective monitoring stack, teams risk missing early signals of system failures, degraded performance, or inefficient resource usage.

What is Prometheus Stack on Kubernetes?

Prometheus and Grafana form the Prometheus Stack, which is the industry-standard observability solution for Kubernetes.

Prometheus handles:

• Metrics collection and time-series storage
• Alerting through Alertmanager
• Automatic target discovery in dynamic environments like Kubernetes

Grafana provides:

• Custom dashboards and visualizations
• Alerts and anomaly detection
• Multi-source data aggregation, including Prometheus, Loki, and SQL databases

Together, they give DevOps teams complete observability of Kubernetes environments.

The key components of Prometheus are:

Prometheus Server

Prometheus Server is the core component of Prometheus that handles metric collection, storage, and serving of time-series data. It pulls metrics from configured targets at specified intervals, stores them locally or remotely, and provides a powerful query language (PromQL) for data analysis. Think of it as the central brain that manages all your monitoring data.

Prometheus Alertmanager

Prometheus Alertmanager manages alerts sent by the Prometheus server. It handles alert grouping, deduplication, silencing, and routing notifications to various services like email, Slack, or PagerDuty. When specified alert conditions are met, Alertmanager ensures the right people are notified through their preferred channels.

Prometheus Targets

Prometheus Targets represent the endpoints that Prometheus monitors. These could be your applications, services, or infrastructure components that expose metrics at an HTTP endpoint (usually /metrics). Each target provides metrics in the Prometheus format that the server scrapes at regular intervals.

Pushgateway

Pushgateway is a component that allows you to push metrics to Prometheus, rather than having them pulled. It's particularly useful for short-lived jobs or batch processes that may not exist long enough for Prometheus to scrape them. Think of it as a metrics buffer that accepts pushed data and holds it until Prometheus pulls it.

Service Discovery

Service Discovery helps Prometheus automatically find and monitor new targets in dynamic environments like Kubernetes. Instead of manually configuring each target, service discovery can automatically detect new services, pods, or instances and update Prometheus's scrape configurations accordingly. This is crucial in cloud-native environments where services come and go frequently.

Prometheus Exporter

Prometheus Exporters are specialized components that convert metrics from existing systems (that don't natively expose Prometheus metrics) into a format Prometheus can understand. For example, the Node Exporter converts system metrics (CPU, memory, disk usage) into Prometheus metrics, while the MySQL Exporter does the same for MySQL database metrics. Exporters act as bridges between your existing systems and Prometheus.

What is Grafana?

Grafana is an open-source interactive and data-visualization platform developed by Grafana Labs. Grafana helps users to query, visualize, explore metrics, and monitor data no matter where they are stored with the help of insightful and customized dashboards and visualizations. You can connect Grafana dashboards with any data sources like Prometheus, MySQL, Influx DB, ElasticSearch, PostgreSQL, etc. Some key features of Grafana are: 

Custom Dashboards

The Grafana dashboard is a set of one or multiple panels, organized into a single pane of glass which provides an at-a-glance view of all key metrics. These panels are customizable and users can create them using components that query and transform raw metrics data from data sources into charts, graphs, and other visualizations.

Alerts

Grafa allows users to set up alerts that notify them of the specific conditions or anomalies in their data i.e. in logs and metrics.

Data Source Compatibility 

Grafana supports a wide variety of data sources users can store their raw metrics data in time series databases like Prometheus and CloudWatch, logging tools like Loki and Elasticsearch, NoSQL/SQL databases like Postgres, CI/CD tooling like GithHub, and many more.

Monitoring Stack with Devtron

Devtron has been built in a modular fashion where you can install different integrations as per your requirements. For monitoring, the default integration is Grafana, as mentioned above. Once Grafana has been installed, now we need to set up a metrics collector, which would scrape all the metrics and act as a data source for Grafana. In this tutorial, we will set up Prometheus and add it to the Devtron dashboard for fetching application metrics.

Grafana

To use the Grafana dashboard, you need to first install the integration from the Devtron Stack Manager. The beauty of integrations that Devtron brings in is that users don't have to worry about the different tools and their complexities. All the heavy lifting is carried out by Devtron and users just need to interact with the Devtron dashboard. In case, you want to access the Grafana dashboard, you need to port forward the devtron-grafana service if Devtron is running in the local system or exposes it via ingress.

kubectl -n devtroncd port-forward svc/devtron-grafana 3000:80 &

Go to your browser and type the URL, In our case, it is localhost:3000, and you should see this. Click on the sign-in button at the bottom left corner.

You should see this on the login page of Grafana.

To login into the Grafana dashboard, your username is admin and you can get your password by executing the following command:

kubectl get secrets -ndevtroncd devtron-grafana-cred-secret -ojsonpath='{.data.admin-password}'|base64 -d

Installing the Prometheus required CRDs

Before we can go ahead and install Prometheus from the chart store, we first need to install the Custom Resource Definitions that are needed for running Prometheus. These CRDs do not get applied by the helm chart, hence we have to do it manually.

The manifest files that we will require for applying CRDs exist within the kube-prometheus GitHub repository. There are multiple manifest files that have to be applied. So let’s first go ahead and clone the repository. Run the below command to clone the repository to your local system.

git clone https://github.com/prometheus-operator/kube-prometheus && cd kube-prometheus

The manifests that we require exist within the manifests/setup directory. However, simply applying these manifests will throw an error. These manifests have to be applied on the server side. We can do this by passing the --server-side flag within the command. Run the below command to apply the CRDs.

kubectl apply --server-side -f manifests/setup

Alternatively, we can apply the CRDs directly from the raw files which exist within the same repository. If you wish to do that, run the below command:

kubectl apply --server-side -f https://raw.githubusercontent.com/prometheus-operator/kube-prometheus/main/manifests/setup/0alertmanagerConfigCustomResourceDefinition.yaml
kubectl apply --server-side -f https://raw.githubusercontent.com/prometheus-operator/kube-prometheus/main/manifests/setup/0alertmanagerCustomResourceDefinition.yaml
kubectl apply --server-side -f https://raw.githubusercontent.com/prometheus-operator/kube-prometheus/main/manifests/setup/0podmonitorCustomResourceDefinition.yaml
kubectl apply --server-side -f https://raw.githubusercontent.com/prometheus-operator/kube-prometheus/main/manifests/setup/0probeCustomResourceDefinition.yaml
kubectl apply --server-side -f https://raw.githubusercontent.com/prometheus-operator/kube-prometheus/main/manifests/setup/0prometheusCustomResourceDefinition.yaml
kubectl apply --server-side -f https://raw.githubusercontent.com/prometheus-operator/kube-prometheus/main/manifests/setup/0prometheusagentCustomResourceDefinition.yaml
kubectl apply --server-side -f https://raw.githubusercontent.com/prometheus-operator/kube-prometheus/main/manifests/setup/0prometheusruleCustomResourceDefinition.yaml
kubectl apply --server-side -f https://raw.githubusercontent.com/prometheus-operator/kube-prometheus/main/manifests/setup/0prometheusruleCustomResourceDefinition.yaml
kubectl apply --server-side -f https://raw.githubusercontent.com/prometheus-operator/kube-prometheus/main/manifests/setup/0scrapeconfigCustomResourceDefinition.yaml
kubectl apply --server-side -f https://raw.githubusercontent.com/prometheus-operator/kube-prometheus/main/manifests/setup/0servicemonitorCustomResourceDefinition.yaml
kubectl apply --server-side -f https://raw.githubusercontent.com/prometheus-operator/kube-prometheus/main/manifests/setup/0thanosrulerCustomResourceDefinition.yaml
kubectl apply --server-side -f https://raw.githubusercontent.com/prometheus-operator/kube-prometheus/main/manifests/setup/namespace.yaml

Now that we have the CRDs applied within the cluster, we can go ahead and install Prometheus

Note: Please make sure that you apply the CRDs before installing the kube-promethus stack. Otherwise, certain components may not be created.

Installing Prometheus

Go to the chart store and search for Prometheus. Use the Prometheus community's kube-prometheus-stack chart to deploy Prometheus. To learn more about this chart, check out the official chart page.

Once you select a chart, you must configure the values per your requirements before deploying. In our case, let's make the following changes.

kube-state-metrics: 
	metricLabelsAllowlist:   
	- pods=[*]
    
serviceMonitorSelectorNilUsesHelmValues: false

podMonitorSelectorNilUsesHelmValues: false

Search for the above parameters and set these values as defined above. Additionally, if you want to expose the Prometheus dashboard via ingress, you can enable the ingress and provide the hostname for it, and likewise for Grafana which is being bundled with kube-prometheus-chart

Now, after installing Prometheus, you need to get the endpoint of the Prometheus server. For every helm chart or application deployed at Devtron, you get a resource-grouped view for all Kubernetes resources deployed along with him. To get the endpoint, under Networking go-to service expand the Prometheus server service and you should be able to see the Endpoints along with EndpointSlice as shown below.

Enable Application Metrics

Not to use Prometheus as a data source for Grafana, go to Global Configurations -> Clusters & Environments and for the respective cluster where you have installed the Prometheus chart, add the endpoint as mentioned below and click on Update Cluster.

Once you add the endpoint, you will be able to see the application metrics in the Devtron dashboard for all the applications deployed in the respective cluster irrespective of the environment (ns) it is being deployed.

With this, you will be able to track your application metrics like resource usage, CPU usage, Throughput, and Latency. For Throughput and Latency, you need to enable the application metrics for respective environments from Deployment Template as mentioned below.

Benefits of Using Devtron for Kubernetes Monitoring

Managing the Prometheus Stack on Kubernetes manually can be complex, especially with multiple clusters. Devtron simplifies observability by:

• Automating Prometheus & Grafana setup with minimal YAML
• Providing resource-grouped Kubernetes views for all workloads
• Offering centralized dashboards for throughput, latency, CPU & memory
• Accelerating root-cause analysis with integrated alerts and application metrics

This ensures faster troubleshooting, reduced downtime, and a better developer experience.

Conclusion

With Prometheus Stack on Kubernetes, you gain powerful visibility into your cluster health, applications, and infrastructure. By combining Prometheus, Grafana, and Devtron, teams can:

• Monitor workloads proactively
• Detect anomalies faster
• Optimize resource usage and application performance

If you’re managing Kubernetes at scale, this observability stack is essential for achieving resilience and operational efficiency.

FAQ

Abhinav Dubey

A Kubernetes head with a travelling soul. Abhinav is a highly motivated and a lifelong learner with a passion for open-source software and cloud-native technologies.

Anurag

Co-authors: Abhinav Dubey, Anurag,