Why I stopped reaching for Kafka the moment two services needed to share an event stream.
Notes on distributed systems, networking, and the slow craft of building software.
I’m an engineer with a soft spot for the unglamorous parts of computer systems — the queueing layers, the TCP retransmits, the gnarly clock skew at 2 a.m. This site is a quiet place where I keep notes that didn’t fit anywhere else.
Most of what I write here is for myself a year from now, when I’ve forgotten exactly why I made a decision and want a paper trail.
“The bug is in the part of the code you didn’t think you needed to read.”
Why I stopped reaching for Kafka the moment two services needed to share an event stream.
A short cheat-sheet for the four flavors of back-pressure I always mix up under load.
It worked for nine months. Then it didn’t. Here’s what I’d do differently.
How a one-line debug log added 38 ms p99 to an otherwise quiet service.
An honest beginner’s walkthrough of a real production capture, with the false starts left in.
I keep code at /projects and a slower-moving reading log at /reading. The best way to reach me is by email; the address is on the about page.