El día en que el pipeline tenía la razón y yo no
Un automerge de Renovate, una regla de pnpm que no cedía, cinco repos con links rotos en las alertas y una actualización de Gitea antes del almuerzo.
Blog DevOps
Homelab, Kubernetes, CI/CD y las cosas que se rompieron en el camino
Mi registry estaba sirviendo capas exactamente al doble de su tamaño
Un ImagePullBackOff un lunes, un error de containerd que era exactamente 2x, y una combinación de flags de buildx que yo mismo había copiado en diez repos.
Tres saltos de Talos y cuatro minors de Kubernetes en un solo domingo
Cómo pasé mi clúster Talos de 7 nodos de la 1.11 a la 1.14 y Kubernetes de 1.34 a 1.37 en un día, y las cinco cosas que se rompieron en el camino.
Diez meses de alertas de fsync en etcd, y la alerta siempre tuvo la razón
Arreglé NFS, después qcow2, después el host. La alerta seguía sonando porque debajo de todo había un SSD sin marca que se quedaba pegado.
Un bot con límite de cero repos me hizo fork de todas formas
Un default de Gitea dejó que un usuario bot restringido hiciera fork, y un fork trae sus propios workflows a mis runners de CI. Una línea lo arregló.
Segmenté la red de mi casa por fases, porque hacerlo de un golpe me dañaba la comida
Cómo separé el Wi-Fi de la casa de la VLAN del homelab con un firewall Sophos: ocho días de logs, luego reglas de allow, el drop de último, y lo que todavía se fuga.
Dos apagones y un portátil que se quedó dormido: lo que mi homelab me enseñó este verano
Se fue la luz dos veces en seis semanas y el cluster se quedó abajo las dos. El arreglo del que tanto me enorgullecí en junio no sirvió de nada en agosto.
Tuve un sitio estático en Kubernetes nueve meses. Después lo moví a un bucket.
Cómo xjohnyx.me pasó de un contenedor Nginx de 40 MB desplegado con Flux a un bucket S3 detrás de CloudFront, y por qué esa era la decisión honesta.
The Day My Pipeline Was Right and I Was Wrong
A Renovate automerge, a pnpm supply-chain gate that would not budge, five repos with broken alert links, and a Gitea upgrade before lunch.
My Registry Was Serving Layers at Exactly Twice Their Size
A Monday ImagePullBackOff, a containerd size error that was exactly 2x, and a buildx flag combination I had copied into ten repos.
Three Talos Hops and Four Kubernetes Minors in One Sunday
How I took my 7-node Talos cluster from 1.11 to 1.14 and Kubernetes from 1.34 to 1.37 in a day, and the five things that broke on the way.
Ten Months of etcd fsync Alerts, and the Alert Was Right Every Time
I fixed NFS, then qcow2, then host tuning. The alert kept firing because a no-name SSD was stalling underneath all of it.
A bot with a repo limit of zero forked my repo anyway
A Gitea default let a restricted bot user fork a repo, and a fork can bring its own workflows onto my CI runners. One line fixed it.
Segmenting My Home Network in Phases, Because Flipping It at Once Would Have Broken Dinner
How I split the house Wi-Fi from the homelab VLAN with a Sophos firewall: eight days of logs first, allow rules second, the drop last, and what still leaked.
Two Blackouts and a Laptop That Fell Asleep: What My Homelab Taught Me This Summer
The power went out twice in six weeks and my cluster stayed down both times. The fix I was so proud of in June did nothing in August.
I Ran a Static Site on Kubernetes for Nine Months. Then I Moved It to a Bucket.
How xjohnyx.me went from a 40 MB Nginx container deployed through Flux to an S3 bucket behind CloudFront, and why that was the honest choice.
Running Kubernetes on OCI Free Tier (Part 3): Networking Deep Dive
Cilium eBPF vs Flannel iptables, Cloudflare Tunnels, and Gateway API — a side-by-side comparison of homelab and cloud cluster networking.
Running Kubernetes on OCI Free Tier (Part 2): Multi-Cluster GitOps with Flux
One Git repo, two Kubernetes clusters. How I manage my homelab and OCI cluster with Flux CD, kustomize overlays, and a shared base.
Running Kubernetes on OCI Free Tier (Part 1): Provisioning with Terraform
How I set up a fully functional Kubernetes cluster on Oracle Cloud's Always Free Tier using Terraform — ARM nodes, networking, and all the gotchas.
The 15-Second Timeout: Debugging Envoy's Silent Killer in Kubernetes
How Envoy's default 15-second timeout silently kills requests before your ALB even responds — and the simple fix.
Taming the Kubernetes Hydra: Finding Happiness in Endless Battles
Every Kubernetes problem you solve creates two more. Here's how I learned to stop fighting the hydra and start working with it.
GitHub Actions: Building a Beta Release Workflow That Actually Works
Input validation, release branching, and automated tagging — a practical GitHub Actions workflow for managing beta releases.
The AWS Decommissioning Checklist I Wish I Had
Lessons from tearing down an AWS project — Route53 gotchas, S3 cleanup tricks, and why you should always snapshot first.
AWS Networking Lessons: Transit Gateways, Shared VPCs, and the IRSA Gotcha
Three AWS networking lessons I learned at work — Transit Gateways vs VPC Peering, Shared VPCs with RAM, and why OIDC providers break canary deployments.
Terraform State Management: Migrating Backends Without Losing Your Mind
A practical guide to migrating Terraform state between backends, cleaning up stale resources, and the commands you'll actually need.
kubectl Commands I Actually Use Every Day
The kubectl commands that actually save me time — from context switching to secret decoding, manifest generation, and more.
Infrastructure as Code: Terraform Best Practices for 2024
Master Terraform with battle-tested patterns for managing cloud infrastructure at scale. Learn state management, module design, and security practices.
Optimizing Docker Images with Multi-Stage Builds
Reduce Docker image sizes by up to 90% and improve security using multi-stage builds. Learn the patterns and best practices.
Getting Started with GitOps using Flux CD
Learn how to implement GitOps practices in Kubernetes using Flux CD for automated deployments and configuration management.