Deploying PostgreSQL on Kubernetes

Deploy production-ready PostgreSQL clusters using CloudNativePG operator (v1.28+) with automated failover.

Quick Start

bash

# 1. Install CloudNativePG operator
kubectl apply --server-side -f \
  https://raw.githubusercontent.com/cloudnative-pg/cloudnative-pg/release-1.28/releases/cnpg-1.28.0.yaml

# 2. Wait for operator
kubectl rollout status deployment -n cnpg-system cnpg-controller-manager

# 3. Deploy PostgreSQL cluster
kubectl apply -f - <<EOF
apiVersion: postgresql.cnpg.io/v1
kind: Cluster
metadata:
  name: pg-cluster
spec:
  instances: 3
  storage:
    size: 10Gi
EOF

# 4. Wait for cluster
kubectl wait cluster/pg-cluster --for=condition=Ready --timeout=300s

Operator Installation

Direct Manifest (Recommended)

bash

kubectl apply --server-side -f \
  https://raw.githubusercontent.com/cloudnative-pg/cloudnative-pg/release-1.28/releases/cnpg-1.28.0.yaml

# Verify
kubectl rollout status deployment -n cnpg-system cnpg-controller-manager
kubectl get pods -n cnpg-system

Helm Installation

bash

helm repo add cnpg https://cloudnative-pg.github.io/charts
helm repo update

helm upgrade --install cnpg \
  --namespace cnpg-system \
  --create-namespace \
  cnpg/cloudnative-pg

Namespace-Scoped (Enhanced Security)

bash

helm upgrade --install cnpg \
  --namespace cnpg-system \
  --create-namespace \
  --set config.clusterWide=false \
  cnpg/cloudnative-pg

Cluster Configurations

Development (Single Instance)

yaml

apiVersion: postgresql.cnpg.io/v1
kind: Cluster
metadata:
  name: pg-dev
spec:
  instances: 1
  imageName: ghcr.io/cloudnative-pg/postgresql:17.2
  primaryUpdateStrategy: unsupervised
  storage:
    size: 5Gi
  postgresql:
    parameters:
      max_connections: "100"
      shared_buffers: "256MB"

Production (HA with 3 Replicas)

yaml

apiVersion: postgresql.cnpg.io/v1
kind: Cluster
metadata:
  name: pg-production
spec:
  instances: 3
  imageName: ghcr.io/cloudnative-pg/postgresql:17.2
  primaryUpdateStrategy: unsupervised

  storage:
    storageClass: standard
    size: 100Gi

  resources:
    requests:
      memory: "2Gi"
      cpu: "1"
    limits:
      memory: "4Gi"
      cpu: "2"

  postgresql:
    parameters:
      max_connections: "200"
      shared_buffers: "1GB"
      effective_cache_size: "3GB"
      maintenance_work_mem: "256MB"
      checkpoint_completion_target: "0.9"
      wal_buffers: "16MB"
      default_statistics_target: "100"
      random_page_cost: "1.1"
      effective_io_concurrency: "200"

  affinity:
    podAntiAffinityType: required  # Spread across nodes

  monitoring:
    enablePodMonitor: true

With Bootstrap Database

yaml

apiVersion: postgresql.cnpg.io/v1
kind: Cluster
metadata:
  name: pg-cluster
spec:
  instances: 3
  storage:
    size: 10Gi

  bootstrap:
    initdb:
      database: learnflow
      owner: app_user
      secret:
        name: app-user-secret

Create the secret first:

bash

kubectl create secret generic app-user-secret \
  --from-literal=username=app_user \
  --from-literal=password=$(openssl rand -hex 16)

Connection Secrets

CloudNativePG automatically creates connection secrets:

Secret	Contents
`pg-cluster-app`	App credentials (recommended)
`pg-cluster-superuser`	Superuser credentials

Get Connection String

bash

# Get app credentials
kubectl get secret pg-cluster-app -o jsonpath='{.data.uri}' | base64 -d

# Get superuser credentials (admin tasks only)
kubectl get secret pg-cluster-superuser -o jsonpath='{.data.uri}' | base64 -d

Use in Deployment

yaml

env:
  - name: DATABASE_URL
    valueFrom:
      secretKeyRef:
        name: pg-cluster-app
        key: uri

Service Endpoints

Service	Port	Use
`pg-cluster-rw`	5432	Read-Write (primary)
`pg-cluster-ro`	5432	Read-Only (replicas)
`pg-cluster-r`	5432	Any instance

Connect from Another Namespace

yaml

env:
  - name: DATABASE_URL
    value: "postgresql://app_user:password@pg-cluster-rw.default.svc.cluster.local:5432/learnflow"

Database Operations

Connect with psql

bash

# Using kubectl cnpg plugin (recommended)
kubectl cnpg psql pg-cluster -- -c "SELECT version();"

# Or directly
kubectl exec -it pg-cluster-1 -- psql -U postgres

Create Database and User

bash

kubectl exec -it pg-cluster-1 -- psql -U postgres <<EOF
CREATE DATABASE myapp;
CREATE USER myapp_user WITH ENCRYPTED PASSWORD 'secure_password';
GRANT ALL PRIVILEGES ON DATABASE myapp TO myapp_user;
\c myapp
GRANT ALL ON SCHEMA public TO myapp_user;
EOF

Run Migrations

bash

# From local machine
kubectl port-forward svc/pg-cluster-rw 5432:5432 &
DATABASE_URL="postgresql://postgres:password@localhost:5432/learnflow" alembic upgrade head

Backup Configuration

Backup to S3

yaml

apiVersion: postgresql.cnpg.io/v1
kind: Cluster
metadata:
  name: pg-cluster
spec:
  instances: 3
  storage:
    size: 10Gi

  backup:
    barmanObjectStore:
      destinationPath: "s3://my-bucket/pg-backups"
      s3Credentials:
        accessKeyId:
          name: s3-creds
          key: ACCESS_KEY_ID
        secretAccessKey:
          name: s3-creds
          key: SECRET_ACCESS_KEY
      wal:
        compression: gzip
      data:
        compression: gzip
    retentionPolicy: "30d"

Schedule Backups

yaml

apiVersion: postgresql.cnpg.io/v1
kind: ScheduledBackup
metadata:
  name: pg-backup-daily
spec:
  schedule: "0 0 * * *"  # Daily at midnight
  backupOwnerReference: cluster
  cluster:
    name: pg-cluster

Monitoring

Check Cluster Status

bash

kubectl get cluster pg-cluster
kubectl describe cluster pg-cluster
kubectl get pods -l cnpg.io/cluster=pg-cluster

View Logs

bash

kubectl logs pg-cluster-1 -f
kubectl logs -l cnpg.io/cluster=pg-cluster --all-containers

Prometheus Metrics

With enablePodMonitor: true, metrics available at:

•cnpg_backends_total - Active connections
•cnpg_pg_replication_lag_seconds - Replica lag
•cnpg_pg_database_size_bytes - Database size

Troubleshooting

Cluster Not Ready

bash

kubectl describe cluster pg-cluster
kubectl get pods -l cnpg.io/cluster=pg-cluster
kubectl logs pg-cluster-1

Connection Issues

bash

# Test connectivity
kubectl run pg-client --rm -it --restart=Never \
  --image=postgres:17 -- \
  psql "postgresql://app_user:password@pg-cluster-rw:5432/learnflow" -c "SELECT 1;"

Common Issues

Error	Cause	Fix
PVC pending	No storage class	Add `storageClass` to spec
Connection refused	Wrong service name	Use `cluster-rw` for writes
Auth failed	Wrong credentials	Check secret `cluster-app`
Replica lag high	Heavy writes	Scale up, increase resources

Cleanup

bash

# Delete cluster (keeps PVCs by default)
kubectl delete cluster pg-cluster

# Delete PVCs (data loss!)
kubectl delete pvc -l cnpg.io/cluster=pg-cluster

# Remove operator
kubectl delete -f \
  https://raw.githubusercontent.com/cloudnative-pg/cloudnative-pg/release-1.28/releases/cnpg-1.28.0.yaml

Verification

Run: python scripts/verify.py

Related Skills

•operating-k8s-local - Local Minikube cluster setup
•scaffolding-fastapi-dapr - FastAPI services with SQLModel
•deploying-kafka-k8s - Kafka for event-driven architecture