Digest #194: What’s Broken in Kubernetes, CI/CD Attacks, FinOps Dashboard and Self-Hosting PostgreSQL.

Kubernetes optimization, CI/CD security, GitHub Actions, Cloudflare Workers, PostgreSQL 18, OpenTelemetry observability, FinOps dashboards, and open-source DevOps tools.

Welcome to this week’s edition of the DevOps Bulletin!

Why Kubernetes clusters are massively underutilized in production, how Cloudflare is eliminating cold starts with “shard and conquer,” and what happened when someone let Claude Code trade $100k for a month. Stack Overflow also shared the story of shutting down its last physical datacenter, Datadog broke down how microservices reshape database usage, and we revisit whether us-east-1 is still AWS’s least reliable region.

On the hands-on side: archiving Git branches as tags, self-hosting Postgres in production, near-instant database clones in PostgreSQL 18, processing S3 access logs at scale, managing GitHub repos with Terraform, building a cost-effective observability stack with OpenTelemetry, and running a production RAG system on Cloudflare for about $5/month.

This week’s video is about CI/CD security + GitHub Actions: how recent supply-chain worms were seeded via Actions weaknesses, and why “AI in your pipeline” is risky.

Open-source picks this week: Kubespray for deploying production Kubernetes clusters across cloud and bare metal, dockcheck to keep self-hosted Docker containers up to date, ultimate-linux as a fun but serious demo of how minimal a Linux distro can be, zoxide as a smarter cd that learns your habits, Databend as an open-source cloud data warehouse on object storage, and watchexec to automatically rerun commands when files change.

FinOps tip of the week: if you want one dashboard that actually helps engineers spot waste, start with AWS CUDOS.

All this and more in this week’s DevOps Bulletin, don’t miss out!

Newsworthy stories

• What’s wrong with Kubernetes today

• Eliminating cold starts 2: shard and conquer

• I gave Claude Code 100k to trade with for a month

• Saying goodbye to the servers at our physical datacenter

• How microservice architectures have shaped the usage of databases

• Is us-east-1 still the least reliable AWS region in 2025?

• Using logs, metrics, traces, and alerts to understand system failures

Tutorials of the week

• Archiving git branches as tags

• Go ahead, self-host Postgres

• Instant database clones with PostgreSQL 18

• S3 server access logs at scale

• Scaling code repository management with Terraform

Enjoying the Bulletin? Consider supporting it with a paid subscription. You’ll keep the free Friday issues and get extras like bonus deep-dives, templates, and the full archive.

• How to build a cost-effective observability platform with OpenTelemetry

• Building production RAG system on CloudFlare for $5/month

• Logging sucks, and here’s how to make it better

• FinOps needs an autopilot

Videos of the week

---

📘 FinOps Tip of the Week

If you want one dashboard that actually helps you spot waste, start with AWS CUDOS. It's built on top of the Cost and Usage Report and answers questions engineers care about.

If you want more hands-on tips like this, check out my latest book, “Practical FinOps”.

---

Projects of the week

• An Ansible-based toolkit to deploy and operate production-ready Kubernetes clusters across cloud and bare metal.

• A lightweight CLI for keeping self-hosted Docker containers up to date without manual work.

• A playful but serious demo of how minimal a Linux distribution can be, built without a traditional libc stack.

• A smarter cd replacement that learns your directory habits and lets you jump anywhere with a few keystrokes.

• An open source cloud data warehouse that unifies analytics, search, and AI on object storage.

• A lightweight CLI that watches files and reruns commands automatically on changes.

Meme of the week

If you have feedback to share or are interested in sponsoring this newsletter, feel free to reach out via LinkedIn or simply reply to this email.