Practical insights on cloud infrastructure, Kubernetes, DevOps, and platform engineering.
Login looped forever after a migration. Four hypotheses, each disproved with direct evidence — the real cause was zero Set-Cookie headers in a 2.7MB browser capture.
Read more →Where AI genuinely sped up a six-week platform build, and the five specific times I had to catch it being confidently wrong.
Read more →A routine TLS ticket turned up zero real certificate infrastructure — just a self-signed issuer that was never meant to face real clients.
Read more →Several tenants failed OTP auth around the same time. One shared cause was tempting. The timestamps didn't agree.
Read more →A VPN flap that should have self-healed instead cascaded into a multi-hour outage — because three factors compounded, not one.
Read more →AWS's dashboard said a VPN outage self-resolved. My own BGP monitoring data said otherwise — and won the argument.
Read more →What we learned operating a fleet of k3s, k8s, and EKS clusters for banking clients — from GitOps workflows to upgrade strategies.
Read more →An async Promise executor anti-pattern turned a routine API failure into an unrecoverable crash, invisible to every try/catch in the codebase.
Read more →ArgoCD self-heal and Kubernetes' garbage collector can race each other and deadlock a Deployment — a runbook for when it happens.
Read more →A vague performance complaint turned up four distinct, real problems — none of which alone would have paged anyone.
Read more →A routine EKS upgrade started crash-looping a service. The upgrade didn't break anything — it removed the slack a bad hardcoded timeout had depended on for years.
Read more →A practical walkthrough of the Terraform and Lambda automation that cut our client's AWS bill through intelligent scheduling.
Read more →How we broke apart a monolithic Terraform codebase into modular, team-owned repositories without breaking production.
Read more →A noisy, zero-action alert couldn't be filtered where it seemed like it should be. The real fix meant understanding the whole automation pipeline's order of operations.
Read more →A GitOps app was stuck on an old image for half a day. Two real bugs turned up along the way — neither explained it. The answer was in last week's commit history.
Read more →A 403 wrapped in a message claiming to be a 500. The real story was a secret that had silently stopped syncing to every pod at once.
Read more →Investigating an out-of-order email trigger turned up something more concerning than the original bug: zero sender validation on a financial transaction trigger.
Read more →