What Is Kubernetes — And When Do You Actually Need It?

TL;DR: Kubernetes is a container orchestrator: it decides which server runs which container, restarts the ones that die, and spreads traffic across the healthy ones. You need it once you're running many services across many machines and deployments have become a manual chore. Below that threshold, it's usually expensive overhead.

Your API went down at 2 a.m. Not a bug — the one server it lived on ran out of memory, and nobody noticed until the support tickets piled up. That's the problem Kubernetes exists to solve. It's also why a lot of teams adopt it years before they need it.

What is Kubernetes, really?

Kubernetes started inside Google as Borg, got open-sourced in 2014, and became the first project hosted by the CNCF. Today it's the default answer to "how do we run containers in production," which is both its strength and its problem.

The distinction people trip on: Docker builds and packages a container. Kubernetes runs containers across a fleet of machines. If Docker is the shipping container, Kubernetes is the port authority — scheduling, routing, replacing broken units, keeping the manifest accurate. You describe the desired state in YAML, and a control loop drags reality toward it, forever.

When to use Kubernetes: real-world examples

Here's what actually justifies the complexity:

• Many services, many teams. Thirty microservices owned by four teams, deploying several times a day. Rolling updates and per-service health checks stop one bad release from taking down everything.
• Traffic you can't predict. An e-commerce shop during a flash sale — the Horizontal Pod Autoscaler adds replicas as CPU climbs, then drops them. On a fixed fleet of VMs you'd be paying for peak capacity all year.
• Portability that matters. The same manifests run on EKS, GKE, AKS, or your own racks. Genuinely useful if you're hybrid or worried about lock-in.
• Batch and ML workloads. Jobs and CronJobs handle training runs and nightly ETL without you babysitting crontab entries on a VM.

When Kubernetes is the wrong answer

Most teams don't need it. I'll say that plainly, because the vendor blogs won't.

• One monolith plus a database → Docker Compose on a $20 VM, or a PaaS like Render or Fly.io. You'll ship faster.
• A small team with no platform experience. The surface area — ingress controllers, cert-manager, RBAC, Helm, CNI plugins, etcd backups — will quietly eat a day a week.
• AWS-only with simple containers → ECS Fargate or App Runner. Fewer moving parts, and no new dialect to learn.
• Static sites and simple APIs → Vercel, Cloudflare Pages, or plain Lambda.

The honest test: if you can't name three things about your infrastructure that hurt this quarter, Kubernetes won't fix them. It'll just relocate the pain.

The middle path: managed Kubernetes

If you do need orchestration, don't run the control plane yourself. AWS EKS charges around $0.10 per hour per cluster for it (roughly $73 a month before a single workload), AKS has a free control-plane tier, and GKE Autopilot bills per pod instead. For edge boxes and homelabs, k3s packs the whole thing into a single binary under 100MB.

Managed isn't the same as hands-off, though. You still own ingress, TLS, autoscaling policy, and — the one nobody warns you about — the cloud bill. Kubernetes makes it easy to request capacity and hard to see where it went. Set cost labels on day one.

Key Takeaways

• Kubernetes schedules, heals, and scales containers across many machines — it's an orchestrator, not a container runtime.
• Adopt it when service count, team count, or traffic volatility outgrows manual deploys.
• For one or two services, a PaaS or Compose setup beats it on cost and speed every time.
• If you go ahead, use a managed control plane and label everything for cost tracking.

Frequently Asked Questions

Do I need Kubernetes if I already use Docker? No. Docker builds and runs containers on one machine. You only need orchestration once one machine isn't enough, or when deployments across several services get messy.

Is Kubernetes hard to learn? The basics take a weekend. Production readiness — networking, storage, security, upgrades — takes months, which is exactly why managed services exist.

What's the simplest alternative to Kubernetes? Docker Compose on a single VM, or a platform like Fly.io, Render, or AWS ECS if you're already on AWS. Start there and migrate when you feel real pain.