# monitoring

A 80k-star, GPL-3.0 monitoring agent that collects every metric every second, runs machine learning on your own hardware, and keeps your telemetry on your infrastructure. Here is the honest deep-dive: architecture, real self-hosting cost, data sovereignty, and where it falls short.

VictoriaMetrics is a 17k-star, Apache-2.0 time series database and monitoring platform written in Go. This deep dive covers its single-binary architecture, MetricsQL, the community-versus-enterprise split, what high cardinality actually costs, and where your metrics go when you own the store.

InfluxDB 3 Core is a 31k-star, MIT/Apache-2.0 time-series engine written in Rust on the Arrow/DataFusion/Parquet stack. A 5,000-word teardown of the v3 rewrite, the diskless object-storage architecture, line-protocol ingest, the built-in Python processing engine, the hard split between open-source Core and closed-source Enterprise, and exactly where your telemetry lands.

Grafana is the AGPL-3.0 observability platform with 76,000+ stars that queries 150+ data sources from one interface. Here's the self-host walkthrough, the provisioning-as-code workflow that stops dashboard rot, the real costs, and why 'Grafana is free' is only half true.

changedetection.io is the Apache-2.0, self-hosted web page change monitor with 33,000+ stars. Here's how it works, why your watchlist is a privacy liability on hosted services, the real costs, and the security defaults you must fix before exposing it.

Uptime Kuma is the MIT-licensed, single-container uptime monitor with 89,000+ stars and 90+ alert channels. Here's how it works, what it really costs, where your monitoring data lands, and the single-instance limits you must plan around.