Hands-on Azure Linux 3.0

No matter if you're doing daily cluster operations or shipping code to AKS, this book turns Azure Linux into a reliable everyday platform.

Microsoft's got a growing number of Azure Kubernetes Service clusters, and they use Azure Linux 3.0 for their own cloud. This book gives programmers, cloud engineers and software teams a practical way to run real workloads on it with confidence. In this hands-on book, we'll take you through building and operating a realistic field service platform with a Python API, a Go worker, PostgreSQL and a ML forecast service. Whatever you change will live in Bicep, Helm and Git, so you can always reproduce the environment and it'll be easy to review.

This book will practically teach you to use Azure Linux clusters with Bicep and Entra ID, and solve problems with nodes using tdnf, logs and kernel tuning. It will show you to use Azure Container Registry, workload identity, Key Vault, network policies and FIPS node pools. You get to see how to secure your workloads, and then scale them with KEDA. In the later chapters, you'll be looking at things like safe patching and upgrades, KQL monitoring, GitOps delivery with GitHub Actions and Flux, GPU training with MLflow, and production backup and cost control.

Key Learnings

Deploy reproducible Azure Linux 3.0 clusters on AKS with Bicep and Entra ID.

Resolve Azure Linux nodes using debug sessions, tdnf, kubelet logs and kernel tuning.

Build slim distroless images, scan them with Trivy and release them through Helm.

Secure workloads with workload identity, Key Vault, Pod Security and FIPS node pools.

Route and restrict traffic using Azure CNI Overlay, Gateway API and Calico policies.

Scale apps and clusters with HPA, KEDA, cluster autoscaler, Spot and zonal pools.

Patch and upgrade node images through upgrade channels and maintenance windows.

Monitor and troubleshoot clusters using Container Insights, KQL and alerts.

Automate delivery with GitHub Actions, OIDC federation and Flux GitOps.

Train, track and serve GPU models, and harden the production with backups.