Thanos

Prometheus covers the short term picture, and Thanos extends it with long term storage, a global query view and highly available setups across many clusters. A sidecar ships samples into object storage, the query layer federates and deduplicates those blocks, and optional components handle downsampling, compaction and alerting fleet wide.

Platform teams adopt it when one Prometheus is no longer enough, for example to keep a year of metrics for every cluster in one place.

Deploying its components in your own stack keeps historical infrastructure metrics under your control, without ingest fees and with a retention period you choose.