Why the order matters

DevOps tools sit on top of each other, so skipping a layer means the next layer never makes sense. Docker is a Linux process with namespaces and cgroups. Kubernetes schedules containers across Linux hosts. Terraform writes the AWS resources those hosts run on. Start with Kubernetes and you will spend weeks copying YAML you cannot debug.

The foundation is Linux because that is what production runs on. W3Techs reports Unix-family systems on 92.0 percent of websites whose operating system is known, with Linux the largest share of that group, and Microsoft reports more than 60 percent of customer compute cores on Azure run Linux. Whatever cloud you end up on, the servers are Linux.

The roadmap at a glance

The month numbers assume 10 to 15 hours a week. With 5 hours a week, roughly double each figure. With 30 hours a week you can compress it into four to five months, which is the pace of the DevOps Engineer course.

Months 1 and 2: Linux and networking

You should be able to log into a fresh Linux server and be useful within ten minutes: the shell (cd, ls, grep, find, sed, awk, pipes), permissions and ownership, users and groups, processes and signals, systemd services and journalctl, apt and dnf, disk and LVM basics, and SSH with keys.

Networking sits inside this stage: IP addresses, subnets, gateways, routes, DNS, ports and HTTP. Practise with ping, dig, curl, ss and tcpdump.

Do this on a real Linux install; a used laptop with Ubuntu or a VirtualBox VM is enough. Work through OverTheWire Bandit for command-line fluency and the MIT Missing Semester for shell habits most graduates never pick up. The Linux System Administration course covers this stage over three months on the RHCSA syllabus if you want a trainer.

Month 3: Git and scripting

You need Git well enough that a merge conflict does not scare you: clone, branch, commit, push, pull, rebase and the pull request workflow. Then write Bash scripts that do real things: rotate logs, back up a directory, restart a stopped service. Add enough Python to parse JSON and call an HTTP API.

The Pro Git book, the Bash manual and the Python tutorial are free and correct. Put every script into a GitHub repository from day one so your portfolio builds itself.

Month 4: Docker

Containers are the unit everything after this point works with. Learn images and layers, write a Dockerfile for a small web app, run it with volumes and a network, bring up an app plus database with Docker Compose, and debug with docker logs and docker exec.

The Docker Get Started guide covers this in a few evenings, and Play with Docker is a free browser sandbox if your laptop is weak. Project: containerise the application from Month 3 and push the image to Docker Hub.

Month 5: CI/CD

A pipeline is a script that runs on every push: checkout, build, test, package, publish, deploy. Build it first in GitHub Actions, then rebuild it in Jenkins so you know both a hosted and a self-hosted tool. Learn how secrets are injected and why a pipeline must never print them.

Use the GitHub Actions docs and the Jenkins documentation. Project: a pipeline that builds the Month 4 image, runs a test and pushes it on every commit to main.

Month 6: AWS core

Learn six services deeply rather than sixty shallowly: IAM, EC2, VPC, S3, RDS in a private subnet and an Application Load Balancer, with CloudWatch for metrics and logs.

AWS Skill Builder has free foundational courses and the AWS Free Tier covers the small instances you need. Set a billing alarm on day one. Project: deploy the Month 5 app to EC2 behind an ALB with the database on RDS.

Month 7: Infrastructure as Code with Terraform

Now rebuild Month 6 as code: HCL, providers, resources, variables, the plan and apply cycle, remote state in S3 with locking, modules, drift and import.

HashiCorp's Terraform tutorials are free and current. Deliverable: a repository that stands up the whole Month 6 environment with one terraform apply and removes it with one terraform destroy.

Months 8 and 9: Kubernetes

Kubernetes takes two months because the concept count is high. Start local with minikube or kind: Pods, Deployments, Services, Ingress, ConfigMaps and Secrets, probes, Horizontal Pod Autoscaler and Helm. Then move the same workload to Amazon EKS, where you meet IAM roles for service accounts and node groups.

The Kubernetes tutorials are free and Killercoda gives you disposable clusters in the browser. Keep EKS sessions short because the control plane is not free: create with Terraform, test for an hour, destroy.

Month 10: Observability

The last stage is knowing whether the system is healthy without logging in: metrics, logs and traces, Prometheus scraping and PromQL, Grafana dashboards and alerts, a simple SLO, and CloudWatch alarms on AWS.

Read the Prometheus overview, the Grafana docs and the monitoring chapters of the free Google SRE book. Instrument the app you have carried through every stage and put a dashboard and one alert on it.

Hours per week and how to spend them

A realistic split is 10 to 15 hours a week: a third reading or watching, two thirds typing at a terminal. DevOps does not transfer through videos alone. Fix one broken thing every week; the free SadServers scenarios build the debugging habit interviewers test.

Keep one application moving through every stage. By the end you have one repository with a Dockerfile, a pipeline, Terraform, Kubernetes manifests and a Grafana dashboard: a portfolio an interviewer can open and question.

Common mistakes freshers make

Starting with Kubernetes or a certification and skipping Linux. The most common one; it shows up in interviews as an inability to debug anything below the YAML.

Watching instead of doing. A 40-hour video course watched passively leaves less behind than 10 hours of breaking and fixing a VM.

Collecting tools instead of depth. Twenty tools at demo level are worth less than Linux, Docker, one CI tool, one cloud, Terraform and Kubernetes done properly.

Ignoring networking. Many incidents are DNS, ports, security groups or routing. If you cannot explain why a request fails to reach a container, the rest of the stack cannot help you.

No portfolio. Certificates are not evidence. A public repository with commits over several months is.

Racing. roadmap.sh and similar maps are good checklists, but order matters more than speed.

Where Techcake programmes fit

Techcake's programmes map onto this roadmap directly. The Linux System Administration course covers Months 1 and 2 over three months. The DevOps Engineer course covers Months 1 to 7 in four months: Linux and networking, Docker, CI/CD with Git, Jenkins and GitHub Actions, and AWS with Terraform. The DevOps Engineer Career Program is the full ten-month route: the same four months of training, then six months of project-based training on campus in Kannur through six production projects, ending with Kubernetes on EKS and Terraform with Prometheus and Grafana.

If you already have the fundamentals, the six-month DevOps project-based training is the entry point. Non-CS graduates should also read the Linux to DevOps path. Fees and EMI options are shared on a free counsellor call. There is no job guarantee; career support means resume review, LinkedIn rebuild, recorded mock interviews and referrals through the mentor network.