Technical Blog
Incident write-ups, pipeline patterns, and architecture decisions — written from the terminal, not the whiteboard.
Featured post
A publicly exposed Postgres port, a compromised container, and a CPU graph that told the whole story. A full incident write-up: detection, containment, eradication, and the firewall rule that should never have existed.
A single JMeter host taps out around 2–5K threads. Getting to 100K means treating your load generators like cattle: wave-based scaling, automated CSV slicing, and startup scripts that do all the thinking.
Five microservice repos, each with per-environment workflow copies quietly drifting apart. The cure: a single branch-conditional workflow per repo, running on self-hosted runners with direct access to the target servers.
The exact architecture serving the site you're reading: a private S3 bucket, CloudFront with OAC, ACM TLS, and GoDaddy DNS pointed at the edge. Global, fast, and roughly the cost of a coffee per year.