Kubernetes Infrastructure Patterns

Infrastructure provisioning using Kubernetes operators, Helm, and GitOps practices.

Core Stack

Component Tool Purpose

Container Docker Image building

Orchestration Kubernetes Workload management

Package Manager Helm Chart management

GitOps ArgoCD, Kustomize Declarative deployments

CI/CD GitHub Actions, Argo Workflows Automation

Monitoring Prometheus, Grafana, Loki Observability

Networking Ingress, NetworkPolicies Traffic management

Context7 Library IDs

Query these for current best practices:

• ArgoCD: /argoproj/argo-cd

• Helm: /helm/helm

Execution Rules

• GitOps first. All changes through git, not kubectl apply ad-hoc

• Helm best practices. Values.yaml for configuration, templates for logic

• Security. No secrets in code, use External Secrets Operator

• Idempotent. All operations safe to retry

• Validate. Always helm template and kubectl diff before apply

Available Operators

Type Operator CRD Kind Namespace

PostgreSQL CloudNative-PG Cluster

databases

Redis/Valkey Redis Operator Redis

databases

S3/Storage SeaweedFS Helm seaweedfs

Kafka Strimzi Kafka

kafka

MongoDB Percona PerconaServerMongoDB

databases

MySQL Percona PerconaXtraDBCluster

databases

NATS NATS Helm Helm nats

RabbitMQ RabbitMQ Operator RabbitmqCluster

messaging

Size Presets

Size CPU Request Memory Storage Replicas

small 100m 256Mi 5Gi 1

medium 500m 1Gi 20Gi 1-2

large 1000m 4Gi 100Gi 3

Infrastructure Provisioning Process

Step 1: Parse Requirements

Extract infrastructure from task XML:

<infrastructure> <component type="postgresql" name="app-db"> <size>small</size> <replicas>1</replicas> <database>app_production</database> </component> </infrastructure>

Step 2: Generate Manifests

Create manifests in the infra/ directory:

infra/ ├── postgresql/ │ └── cluster.yaml ├── valkey/ │ └── redis.yaml ├── seaweedfs/ │ └── bucket-init.yaml └── kustomization.yaml

Step 3: PostgreSQL Example

apiVersion: postgresql.cnpg.io/v1 kind: Cluster metadata: name: app-db namespace: databases spec: instances: 1 storage: size: 5Gi storageClass: mayastor bootstrap: initdb: database: app_production owner: app_user

Step 4: Valkey/Redis Example

apiVersion: redis.redis.opstreelabs.in/v1beta2 kind: Redis metadata: name: app-cache namespace: databases spec: kubernetesConfig: image: redis:7-alpine storage: volumeClaimTemplate: spec: storageClassName: mayastor accessModes: ["ReadWriteOnce"] resources: requests: storage: 1Gi

Step 5: Apply and Wait

Apply manifests

kubectl apply -k infra/

Wait for PostgreSQL

kubectl wait --for=condition=Ready cluster/app-db -n databases --timeout=300s

Wait for Valkey

kubectl wait --for=condition=Ready redis/app-cache -n databases --timeout=300s

Step 6: Create Infrastructure ConfigMap

apiVersion: v1 kind: ConfigMap metadata: name: app-infra-config namespace: app labels: cto.platform/type: infrastructure-config data: DATABASE_URL: postgresql://app_user:$DB_PASSWORD@app-db-rw.databases.svc:5432/app DATABASE_HOST: app-db-rw.databases.svc DATABASE_PORT: "5432" DATABASE_NAME: app

REDIS_URL: redis://app-cache.databases.svc:6379 REDIS_HOST: app-cache.databases.svc REDIS_PORT: "6379"

S3_ENDPOINT: http://seaweedfs-filer.seaweedfs.svc:8333 S3_BUCKET: app-uploads

Validation Commands

Helm validation

helm lint ./chart helm template ./chart --debug

Kubernetes validation

kubectl diff -f manifest.yaml kubeval manifest.yaml

ArgoCD

argocd app diff app-name

Check status

kubectl get all -n databases kubectl get cluster -n databases -o wide kubectl get redis -n databases

Error Handling

If provisioning fails:

• Check operator logs: kubectl logs -n operators -l app.kubernetes.io/name=<operator>

• Describe the resource: kubectl describe cluster/app-db -n databases

• Check events: kubectl get events -n databases --sort-by='.lastTimestamp'

• Verify storage class: kubectl get storageclass mayastor

Guidelines

• Use operators for stateful services (databases, caches)

• Store connection details in ConfigMaps for other agents

• Always wait for resources to be ready before completing

• Document connection information in infra/README.md

• Use GitOps (ArgoCD) for production deployments

• Never hardcode secrets in manifests