Scriber

A small, self-contained service that records operational events as time intervals—update runs, backup windows, maintenance jobs—in a single local SQLite file, and offers them to Grafana as annotations.

The gap it closes is causality: metrics tell you that I/O load spiked, the annotation tells you which backup caused it. Every job reports its start, and later its end; the service correlates both into one interval with duration, tags, and outcome. A job that dies without reporting its end is closed as aborted after a configurable timeout instead of growing forever.

Deliberately unambitious about infrastructure: one binary, one database file, no cluster, no external runtime. Events are sparse and long-lived, which is exactly what a metric store handles worst—so Scriber keeps them separate, readable, and independent of monitoring retention.