Zero-Downtime Rollbacks Are a Deployment Design Problem
A rollback is only as fast as the slowest thing it has to undo. Most teams discover this at 2am, mid-incident, with a database migration in the way.
Library
Posts
Everything we have published, newest first.
12 posts
Your Dockerfile Is Probably Three Times Bigger Than It Needs to Be
Layer order, multi-stage builds and the cache behaviours nobody explains until your CI queue is forty minutes deep.
Terraform State Is a Database, So Stop Treating It Like a Log File
Remote state, locking and the drift problem, framed the way you would frame any other shared mutable store.
The Four Golden Signals, and the Two Everyone Forgets
Latency, traffic, errors and saturation are the canonical four. The pair that actually saves you at 3am are the ones about saturation and ti…
We Cut CI From 22 Minutes to 6, and Most of It Was Waiting
A teardown of a slow pipeline: what the wall-clock was actually spent on, and the four changes that bought back sixteen minutes.
cgroups From First Principles: Why Your Container Has No Idea How Big It Is
Containers are not virtual machines. They are a set of kernel bookkeeping rules, and most "container memory" bugs are really cgroup surprise…
Your Base Image Is a Supply Chain Decision, Not a Convenience
Every FROM line is a dependency you never version-pin and rarely scan. A practical model for choosing and maintaining base images.
Spot Instances Are Cheap Until the Wrong Job Lands on One
A 70% discount with a two-minute eviction notice is a great deal for some workloads and a terrible one for others. How to tell which is whic…
Health Checks That Lie: liveness, readiness and the Restart Loop
A liveness probe that checks the database turns a slow dependency into a cluster-wide restart storm. Here is the discipline that prevents it…
Keep exploring