A practical AWS DevOps path: CI/CD, IaC, and EKS without getting lost
AWS has the pieces for DevOps, but they are scattered. This article maps a hands-on path from first pipeline to an app running on Amazon EKS, with the exact services, projects, and order I would follow as a beginner.
A practical AWS DevOps path: CI/CD, IaC, and EKS without getting lost
AWS already has the tools to practice DevOps. The hard part is order.
Most beginners open the console, see CodePipeline, CDK, EKS, CloudWatch, IAM, and try to learn all of them at once. That usually ends with a few tutorials and no system they can explain in an interview.
DevOps on AWS is one loop:
code → build → infrastructure → deploy → observe → improve
If you learn that loop with one small app, the services start to make sense.
What to ignore at the start
Skip these until you have one working path:
- Every AWS service on the homepage
- Multi-account landing zones
- Service Mesh
- GitOps with a full platform team
- Trying EKS before you can ship a container
You need Linux basics, Git, Docker, IAM, and a VPC mental model. That is enough to start.
The path
1. One app, one container
Pick a tiny API or web app. Wrap it in Docker. Run it locally.
Then push the image to Amazon ECR.
If you cannot build and run the container on your laptop, do not open EKS yet.
2. CI before CD
Create a pipeline that runs on every push:
- lint / unit test
- build the image
- push to ECR
Use GitHub Actions or AWS CodePipeline + CodeBuild. Either is fine. What matters is that a git push produces an artifact without you clicking around the console.
3. Infrastructure as code, not click-ops
Use AWS CDK or Terraform. CloudFormation is acceptable, but CDK/Terraform is faster to learn for most people.
First stack should be boring:
- VPC with public/private subnets
- IAM roles with least privilege
- ECR repository
- security groups that are not
0.0.0.0/0on every port
If you cannot destroy and recreate the stack, it is not infrastructure as code yet.
4. Deploy somewhere simpler than EKS first
Before Kubernetes, deploy the same image to Amazon ECS on Fargate.
Why: you practice CI/CD, networking, IAM, and rollouts without also learning etcd, nodes, and kubectl at the same time.
When that works, move the same app to Amazon EKS.
5. EKS as a target, not a personality
On EKS, keep the first version small:
- managed node group or Fargate profile
- one Deployment + Service + Ingress or ALB controller
- rolling update
- a known way to roll back
You are done with this stage when a pipeline can ship a new image tag to the cluster.
6. Operate it
Add the part most tutorials skip:
- CloudWatch logs and metrics
- an alarm on 5xx or failed pipeline
- IAM review: what can this role do if the token leaks?
A pipeline that deploys but tells you nothing when it breaks is not DevOps. It is automation with hope.
Three projects that are enough
- Docker + CI: app builds and lands in ECR on every push.
- IaC + ECS/Fargate: VPC + service defined in code, pipeline deploys it.
- Same app on EKS: pipeline rolls out and can roll back. Alarms exist.
Put all three in public repos with a README that explains the flow, not only the commands.
Certifications, in the right order
Certs help after you have projects, not before.
- Cloud Practitioner only if you are brand new
- Solutions Architect Associate
- DevOps Engineer Professional
A repo with a real pipeline beats a badge with no artifacts.
What good looks like after 8–12 weeks
You can tell a hiring manager:
- how a commit becomes a running container
- why the app is in a private subnet
- how IAM is scoped
- how you deploy and how you undo a bad release
- what you watch in CloudWatch
That is a DevOps beginner who can grow. That is better than someone who “knows EKS” but cannot ship.
A request to AWS
Builder Center already has strong workshops on CodePipeline, CDK, and EKS. What is missing is one ordered path that connects them.
If AWS ships a single hands-on DevOps path with a sample repo at the end, fewer builders will get stuck collecting services instead of learning the workflow.
Until then, build the path yourself. Keep the app small. Keep the loop complete.
Enjoyed reading this content? Let the author know!
Your likes, comments, shares, and saves help creators reach more builders.
Loading recommendations
Loading article