Deploy the prometheus-haproxy-exporter helm chart on Kubernetes

Question

Pulumi · Accepted Answer

Deploying the Prometheus HAProxy Exporter using a Helm chart on a Kubernetes cluster is a common way to monitor HAProxy instances. We will use Pulumi's Kubernetes package to accomplish this. Specifically, we'll leverage the Chart resource, which allows us to deploy a Helm chart.

Here's how you can do it using Pulumi with TypeScript:

We'll first import the @pulumi/kubernetes package, which provides the Kubernetes provider and necessary resources.
We'll then use the Chart resource to deploy the Prometheus HAProxy Exporter Helm chart.
We'll need to specify the repository where the chart is located and the chart name.
Optionally, you can provide values to customize the deployment according to your needs, such as setting the HAProxy instance to monitor.

Let's walk through the Pulumi TypeScript program that performs this deployment:

import * as k8s from "@pulumi/kubernetes";

// Create a Kubernetes Chart resource using the prometheus-haproxy-exporter chart from the Helm repository.
const prometheusHaproxyExporterChart = new k8s.helm.v3.Chart("prometheus-haproxy-exporter", {
    // Specify the repository and the chart name.
    chart: "prometheus-haproxy-exporter",
    repo: "https://prometheus-community.github.io/helm-charts",
    // Specify the version of the chart you'd like to deploy.
    version: "1.0.0",
    // Replace with the namespace where you want to deploy the exporter.
    namespace: "monitoring",
    // Provide custom values to the Helm chart. For instance, if you need to set up the address of your HAProxy instance.
    values: {
        // Use this section to input your custom values according to the chart's expectations.
        // This is a generic placeholder; you will need to replace it with actual values.
        serviceMonitor: {
          enabled: true, // Enables a ServiceMonitor for the Prometheus Operator.
        },
        // Configure the exporter with the address of HAProxy:
        haproxy: {
          // Define the address and port of the HAProxy instance you wish to monitor.
          // For example: "http://<haproxy_address>:<haproxy_port>"
          address: "http://haproxy-exporter:9101"
        }
    }
});

// Export the address of the HAProxy Exporter service, which can be used to access the metrics endpoint.
export const exporterServiceAddress = prometheusHaproxyExporterChart.getResourceProperty("v1/Service", "prometheus-haproxy-exporter", "status").apply(status => 
    status.loadBalancer.ingress[0].hostname || status.loadBalancer.ingress[0].ip
);

This program defines a Helm chart resource for the Prometheus HAProxy Exporter. You need to install the @pulumi/kubernetes SDK using npm if you haven't already (npm install @pulumi/kubernetes).

The code above assumes you have an existing Kubernetes cluster and Prometheus set up. If any of these prerequisites are not met, you will need additional Pulumi resources to create those.

Let's break down some of the code:

The new k8s.helm.v3.Chart constructor is used to deploy the Helm chart to Kubernetes. Replace "https://prometheus-community.github.io/helm-charts" with the actual repository URL if different and 1.0.0 with the desired chart version.
You may need to specify the target namespace where the chart should be deployed with namespace: "monitoring". Ensure this namespace exists in your cluster.
values are the configuration parameters that are specific to this chart. In a real deployment, you'd adjust these to fit your particular monitoring needs.
Lastly, we export the endpoint address using export const exporterServiceAddress. If the service is of type LoadBalancer, and your cloud provider supports it, you will get an externally accessible IP or hostname.

After you've written this code in a file named index.ts, you can deploy it using Pulumi CLI commands:

pulumi up to preview and deploy the changes.
Review the proposed changes and approve them to start the deployment.

Finally, check that the Prometheus HAProxy Exporter is running by observing the outputted service address or by using kubectl commands to see the deployed resources.