DevOps

Module 1: Linux, networking, Git

Foundations

Six months of DevOps starts on a box you can SSH into. You will read logs, permissions, and a packet path before you touch Kubernetes.

Linux

• Users, processes, systemd, disks
• ssh, scp, and why root login is a smell
• A broken permission you fix without Stack Overflow first

Networking enough

• DNS, ports, TLS at a practical level
• curl, ss, and ‘connection refused’ vs timeout
• Firewalls as a product decision

Git and PRs

• Branch, rebase-or-merge as a team rule
• Reviews that catch a secret
• Tags for releases

Not this module

• Not ML training pipelines
• Not Power BI
• Not a cloud certification dump on day one

You leave with a Linux lab note: users, a service, and a git history that is readable.

Module 2: CI/CD

Foundations

If it is not in the pipeline, it is not a release. You will build, test, and promote with a gate that can fail.

Pipeline design

• Build → test → artifact → deploy
• Environments: dev/stage/prod as config, not hope
• Secrets in the runner, not in YAML

GitHub Actions-class

• Workflows, caches, matrix builds
• Required checks on the default branch
• A failed job you debug from logs

Artifacts

• Versioned images or packages
• SBOM-lite thinking
• Rollback is a previous artifact, not a prayer

Culture

• You do not hotfix prod from a laptop as the process
• A runbook: how to revert in 15 minutes

You leave with a CI workflow that builds and fails closed on tests.

Module 3: Docker

Containers and cloud

Images are how you stop ‘works on my machine’. You will write Dockerfiles that are boring and reproducible.

Images

• Multi-stage builds, non-root, .dockerignore
• Pin tags; latest is an incident
• Scan-lite: know what you shipped

Compose

• App + db locally
• Healthchecks
• The same entrypoint CI will use

Registries

• Push/pull, tags as versions
• Never bake secrets

Handoff to k8s

• A container that probes can understand
• Resource needs written down

You leave with a Dockerfile and compose file that run from a cold clone.

Module 4: Kubernetes and IaC

Containers and cloud

Kubernetes is a workload API. Terraform (or equivalent) is how the cluster and cloud bits stop being click-ops.

Kubernetes

• Deployments, services, probes, ConfigMaps/Secrets
• Requests/limits
• A rollout and a rollback

Ingress and DNS

• How traffic enters
• TLS at the edge
• A 502 you can explain

Terraform / IaC

• State, plans, and why apply is a review
• Modules without a 4,000-line main
• Destroy is a skill

Cloud

• One provider enough to be honest (IAM, networks, a cluster or VMs)
• Cost as a dashboard item
• This is still not MLOps

You leave with manifests plus an IaC sketch that creates something you can destroy.

Module 5: Observe and operate

Containers and cloud

Ops dashboards here are Prometheus/Grafana-style. You will page on symptoms, not vibes. Distinct from MLOps drift jobs.

Observability

• Logs, metrics, traces at a practical trio
• SLOs: latency and error rate
• Alerts a human will not mute

Incidents

• A runbook: deploy went bad
• On-call notes without hero culture as the product
• Postmortem: blameless, one action

Security-lite

• Least privilege IAM
• Image and secret hygiene
• Audit: who applied what

Career call

• DevOps vs MLOps vs LLMOps — pick one gate
• Fee ₹60,000 / 6 months as published

You leave with a Grafana-style board (or substitute) and an incident runbook you tested.