flochaiWhat we would tell a friend
who runs a platform.
Short reads on hiring, deploys and risk, written the way we explain them across a table. No jargon for its own sake, trade-offs left in.
Debug Detective: the locked library
My pipeline could not read the components it was including from another project. The file was there. The path was right. The error was lying to me on purpose, and the real answer was in a layer I had not thought to suspect.
Read the articleNobody was overspending on purpose
One managed-Kafka line item was returning about $49,000 a year. A copy-pasted resource block was holding 31 vCPU across three clusters. Why platform findings of that size usually cover the engineer who finds them, and when they do not.
Read the article
Debug Detective: the phantom rule
Apache would not start, something else had taken its port, and killing it fixed nothing. The firewall rule that should have let me through was letting nobody through, and the reason was in a column I was not printing.
Read the articleThe migration that failed for four months
A self-managed GitLab carried a background migration failing since April. Diagnosing it down to an upstream bug, clearing it from the Rails console, and writing the DR runbook that should have existed all along.
Read the articleWhat a DevOps hire really costs
Salary is the small part: months of searching, one person covering five specialties, and a bus factor of one. The real math of DevOps hiring, and the alternative.
Read the articleHow to stop fearing your deploys
Deploy fear is a process smell, not a personality trait. What a boring pipeline looks like: gates, health checks, one-step rollbacks, and runbooks that make 3am survivable.
Read the articleThe bus factor: when your platform rides on one engineer
One person holds the infrastructure, everyone else hopes. How platforms end up with a bus factor of one, what it costs before anyone leaves, and how to fix it without hiring.
Read the articleDuct tape to trust: our DevSecOps lab, rebuilt
A tiny Streamlit app run through a full DevSecOps pipeline: seven stages, Grype, Playwright in CI, GitOps with Argo CD, poll-based health checks and automatic rollback. What changed between v1 and v2, and why.
Read the article