Kubernetes production deployment: a practical guide

Key takeaway

In one line: Kubernetes is an operational model of control plane + nodes + pods. Without probes, resource limits, and rollout strategy aligned in YAML, you get “deploy succeeded but the service is down” on repeat.

---

Introduction

Kubernetes is the de facto standard for container orchestration, but YAML and concepts have a learning curve. Here is how we set requests/limits, health checks, and rollouts in production.

Basic deployment

Deployment

Service

Resource management

Requests and limits

Best practices

• Requests: minimum resources the scheduler guarantees
• Limits: maximum the container may use
• A 1:2 or 1:4 ratio between requests and limits is a common starting point

ResourceQuota

Health checks

Liveness probe

Confirms the app is alive; failures restart the container.

Readiness probe

Confirms the app can accept traffic; failures remove the pod from Service endpoints.

Startup probe

For slow-starting applications.

Autoscaling

Horizontal Pod Autoscaler (HPA)

Vertical Pod Autoscaler (VPA)

Deployment strategies

Rolling update

Blue-green

To cut over to green, change the Service selector.version to green.

Canary

ConfigMap and Secret

ConfigMap

Secret

Network policies

Monitoring and logging

Prometheus scrape annotations

Log collection (Fluentd)

Security best practices

Pod Security Policy (legacy API)

RBAC

Conclusion

For successful Kubernetes production deployments:

1. Resource discipline: set requests and limits deliberately
2. Probes: use liveness, readiness, and startup probes appropriately
3. Autoscaling: HPA and VPA to match load
4. Safe rollouts: rolling, blue-green, and canary
5. Security: NetworkPolicy, RBAC, and pod hardening
6. Observability: metrics and centralized logs

Combining these patterns yields a stable, scalable cluster.

References

• Kubernetes documentation
• Configuration best practices
• What is Kubernetes