Building in public
Engineering notes from the junkyard.
I'm rebuilding Junkyard, my Go log aggregator, into a personal observability platform: OTLP ingestion, a custom time-series database, Kubernetes deployment, and writing up every design decision along the way.
// Go · OpenTelemetry · Kubernetes · distributed systems, one post at a time
Posts
A spec-compliant OTLP receiver, brought to you by a dead SSD
Windows 11 corrupted my entire SSD six hours of recovery couldn't save. What survived is everything I ever pushed — including the walking skeleton this post was supposed to be about: partial_success, backpressure, and why the queue is just a channel.
Read →Rebuilding Junkyard: a mini-Datadog from scratch
Why I'm throwing away a working log aggregator and rebuilding it as an OTLP-native observability platform, and the architecture I'm starting with: a deliberate ingestion pipeline, a custom TSDB, and a documented reason for every trade-off.
Read →