Uptime monitoring in Go that fits in one binary and shows latency next to state. The UI is embedded in the executable, so there's no nginx alongside it and no static folder to drift out of sync. A personal project of mine, open source under AGPL-3.0.
Case Study
Problem
Almost every incident starts with a service getting slow, not going down, and a monitor that only knows up or down arrives late to that story. Most options also want an nginx alongside and a folder of static files, which makes the monitor more machinery to maintain than the service it watches.
Technical Decision
One binary, with the UI embedded via go:embed. Pluggable storage across SQLite and PostgreSQL, with the same conformance battery running identically on both so pluggable database doesn't become a facade. Raw beats roll up into hourly and daily aggregates, and percentiles always come from the raw data, never a percentile of percentiles. An independent probe separates the target going down from the network going down, and it only earns the power to silence an alert after proving it works, because a probe blocked by a firewall would mute everything forever.
Result
27 thousand lines of Go, 12 thousand of them tests, across 651 test functions. The only surface without credentials, the public status page, takes a battery of intrusion tests with path traversal, injection in the slug, and a forged Host header. Image on GHCR, project site live, AGPL-3.0.
