// the find
cortexproject/cortex
A horizontally scalable, highly available, multi-tenant, long term Prometheus.
Cortex is a distributed, multi-tenant Prometheus backend that adds horizontal scalability, HA replication, and long-term storage (S3, GCS, Azure) to what vanilla Prometheus cannot do alone. It is what you reach for when a single Prometheus instance stops being enough and you need tenant isolation — think managed monitoring platforms or large organizations running Kubernetes at scale.
The blocks storage engine (TSDB-compatible, backed by object storage) means you pay S3 prices for long-term retention instead of running expensive replicated disk. The component model lets you scale ingesters, queriers, and compactors independently — a single binary mode exists for dev but microservices mode is where it earns its keep. The integration test suite is legitimately thorough: real Docker-based end-to-end tests covering zone-aware replication, ruler HA, and OTel ingest. Shuffle sharding limits blast radius so one noisy tenant cannot bring down queries for everyone.
Thanos and Grafana Mimir (Cortex's commercial fork) have both pulled significant engineering attention away from this project — Mimir in particular has diverged enough that blog posts and StackOverflow answers frequently apply to one but not the other, and you will waste time sorting out which. The configuration surface is enormous and the YAML reference doc is the kind of thing you read with a cup of coffee and still miss things; there is no good 'minimal production config' that is kept current. Operationally, the ingester is a stateful set with WAL, so rolling upgrades require careful ordering — the docs cover it, but it is easy to get wrong the first time and corrupt in-flight data.