What is your experience with using Terraform in the deployment process of Helm charts?

I use Terraform to manage the lifecycle of Helm releases, including:

• Infrastructure provisioning: Creating Kubernetes clusters (AWS EKS, Google GKE, Azure AKS), databases, and other necessary resources that will be used by Helm.
• Deploying Helm releases: Using the helm provider in Terraform to install, update, and delete Helm charts in Kubernetes.
• Managing dependencies: Defining dependencies between infrastructure resources and Helm releases, ensuring that the infrastructure is ready before attempting to deploy a chart.
• Managing states: Storing Terraform state, including information about deployed Helm releases, to ensure idempotency and rollback capabilities.
• Version control: Versioning Terraform configuration along with charts and application code.

Example of using the helm provider:

resource "helm_release" "myapp" {
  name       = "my-app"
  repository = "https://charts.example.com/"
  chart      = "my-chart"
  version    = "1.2.3"
  namespace  = "default"

  values = [
    file("values.yaml") // Ability to pass values from a file
  ]

  set { // Or set individual values directly
    name  = "replicaCount"
    value = "3"
  }
}

Advantages of this approach:

• Infrastructure as code: Managing both infrastructure and application deployment from a single source.
• Automation: Full automation of the process from infrastructure creation to Helm release deployment.
• Idempotency: Applying Terraform configuration always results in the desired state.
• Transparency: Easily track which resources and releases are deployed in the cluster.
• Dependency management: Clear control over deployment order.

Despite the advantages, it is important to note that Terraform does not replace Helm charts themselves. Helm remains responsible for packaging and managing applications in Kubernetes, while Terraform integrates this process into a broader IaC strategy.