Prometheus - Intro, Deep Dive, And Open Q+A
summary
Prometheus is a metrics-focused, pull-based system that scales to about 1 million samples/sec on a single node, compresses samples to ~1.2–1.6 bytes each, and typically keeps weeks of data (single-node, a few TB, not multi-year). Recent and upcoming features include remote_write agent mode (scrape-forwarding without local retention), experimental out-of-order sample ingestion (configurable time window), and native histograms in v2.40. The session covers practical limits (scrape/sample caps, relabel rules), LTS releases, PromLens/tooling, and real-world failure modes and mitigations.
AI-written summary · about
chapters (14)
- 0:00Start
- 0:10What is Prometheus?
- 2:32Architecture
- 6:22Selling Points
- 6:53Data Model
- 8:08Querying
- 11:38Efficiency
- 13:05Bridging the gap
- 14:19Conclusion
- 15:03PromQL
- 17:19Agent mode
- 18:00LTS releases
- 19:48What's Coming
- 35:33Remote Write Receiver
