Engineering Notes
Code is abundant now. Architecture, maintainability, and staying true to what the application is for are still the hard part.
Rebuild Staging From Production Backups
Before each release, I rebuild staging from the latest production backup. The release gets tested on real data shapes, and the restore process gets tested too.
How I Recover Stale Crawler Jobs
When a crawler worker dies mid-job, the rows it claimed sit in a started state forever. Here is the timeout, retry count, and alert that fixed that.
Fixing Prometheus High-Cardinality Metrics
Unbounded labels like user and property IDs inflate Prometheus metrics. I removed them, kept bounded dimensions, and fixed the design instead of changing platforms.
Email Connections That Survive Auth Expiry
Email in a multi-tenant app has two readers and several providers with different login lifetimes. This is how I kept sync reliable and what happens after three failures.
Moving Image Processing to the Browser
Converting and resizing photos in the browser, behind an upload queue, kept heavy image work off the server. In testing, 70+ photos went live in under two minutes.