Kubernetes Fundamentals

Introduction

When running applications across dozens of servers, managing containers manually is impossible. Kubernetes (K8s) is an open-source container orchestration engine that automates deploying, scaling, and managing containerized applications at scale.

Background

Unlike Docker Compose, which runs on a single host, Kubernetes manages clusters of virtual or physical machines. The core abstractions are:

1. Pods: The smallest deployable units containing one or more containers.
2. Deployments: Declarative templates that define the desired state for Pod replicas.
3. Services: Network access definitions that expose Pods using stable DNS addresses.

Implementation

We configured a scalable Kubernetes Deployment that runs 3 replicas of our API and exposes them using a load balancer.

Challenges

Running Kubernetes in production reveals typical orchestration challenges:

• Resource Contention: Unchecked containers devouring CPU or memory, causing neighboring Pods to crash (OOM killed).
• Zero-downtime Rollouts: Deploying new images without dropping active connections.

Solutions

We solved this by establishing cluster configurations:

1. Resource Limits: Setting explicit requests and limits (resources: limits: memory: 512Mi, cpu: 500m) for every container.
2. Liveness and Readiness Probes: Configuring health endpoints to ensure Kubernetes only routes traffic to fully initialized Pods during rollouts.

Results

Configuring resource limits stabilized node RAM usage and eliminated all neighboring container OOM crashes, while readiness probes allowed zero-downtime rollouts of API updates.

Conclusion

Kubernetes is highly resilient but requires explicit resource declarations. By setting requests/limits and configuring liveness/readiness health checks, you build self-healing deployments that run reliably.