Infrastructure As Code

Infrastructure as Code

Learn to THINK in declarations, not manual clicks.

⚠️ Core Principles

Declarative

• Define desired state
• Let tool reconcile
• Idempotent operations

Version Controlled

• Git for infrastructure
• Code review for changes
• Audit trail

Common Patterns

Terraform - Kubernetes Cluster

# main.tf
resource "kubernetes_namespace" "data" {
  metadata {
    name = "data-platform"
    labels = {
      environment = var.environment
    }
  }
}

resource "helm_release" "spark_operator" {
  name       = "spark-operator"
  repository = "https://kubeflow.github.io/spark-operator"
  chart      = "spark-operator"
  namespace  = kubernetes_namespace.data.metadata[0].name

  set {
    name  = "webhook.enable"
    value = "true"
  }
}

resource "helm_release" "kafka" {
  name       = "kafka"
  repository = "https://charts.bitnami.com/bitnami"
  chart      = "kafka"
  namespace  = kubernetes_namespace.data.metadata[0].name

  values = [file("${path.module}/values/kafka.yaml")]
}

Pulumi - Python

import pulumi
import pulumi_kubernetes as k8s

namespace = k8s.core.v1.Namespace(
    "data-platform",
    metadata=k8s.meta.v1.ObjectMetaArgs(
        name="data-platform"
    )
)

spark_operator = k8s.helm.v3.Release(
    "spark-operator",
    chart="spark-operator",
    repository_opts=k8s.helm.v3.RepositoryOptsArgs(
        repo="https://kubeflow.github.io/spark-operator"
    ),
    namespace=namespace.metadata.name,
)

Helm - Spark Application

# values.yaml
spark:
  image: spark:3.5
  driver:
    cores: 1
    memory: "2g"
  executor:
    instances: 3
    cores: 2
    memory: "4g"

minio:
  endpoint: "http://minio:9000"
  accessKey: admin
  secretKey: password

GitOps with ArgoCD

# application.yaml
apiVersion: argoproj.io/v1alpha1
kind: Application
metadata:
  name: data-platform
  namespace: argocd
spec:
  project: default
  source:
    repoURL: https://github.com/org/data-platform
    path: k8s/overlays/production
    targetRevision: HEAD
  destination:
    server: https://kubernetes.default.svc
    namespace: data-platform
  syncPolicy:
    automated:
      prune: true
      selfHeal: true

Best Practices

Anti-Patterns

Related Skills

• For K8s workloads: kubernetes-data
• For Docker: docker-data

Converted and distributed by TomeVault — claim your Tome and manage your conversions.