Prometheus Deep Dive
Talkat KubeCon + CloudNativeCon North America 2019 · on Thursday, 2019-11-21· at 10:55· Ballroom Sec 20AB - San Diego Convention Center, Upper Level·
summary
Prometheus is an intentionally uncoordinated system with a local TSDB and right‑ahead log, designed to run close to targets for reliable alerting; it extrapolates sparse scrapes when computing rate/increase which avoids undercounting across offset scrape windows. Capacity planning centers on ingestion: practical example is 100k samples/sec using about 1.5 bytes per sample (≈0.5 GB/hour). Scaling options include federation + recording rules, and long‑term stores like Thanos, Cortex or M3; caching (Trickster) and a new React web UI are in active use.
AI-written summary · about
chapters (17)
- 0:00Intro
- 2:25Prometheus Design
- 4:40Prometheus Data Collection
- 10:10Scaling
- 13:13Service Discovery
- 13:48QA
- 14:27Index
- 16:30Recording Rules
- 17:36Thanos
- 21:03Recommendations
- 22:18Problems with Prometheus
- 23:24Application metrics
- 24:00Web servers
- 25:50Vertical compaction
- 27:10Prometheus web interface
- 28:31TSDB sizing
- 29:13Config management
related talks
35mPushing the Limits of Prometheus at Etsy
16mKeynote: Prometheus 2.0 – The Next Scale of Cloud Native Monitoring
35mPrometheus Sparse High-Resolution Histograms in Action
27mThanos: Highly Available, Pluggable, Long Term Metric Storage for Everyone!
35mStateless Collectors For Stateful Data: Scaling Prometheus As a Node Agent
30mDistributing PromQL for Fast and Efficient Kubernetes Fleet Monitor...
13mLightning Talk: A Drop-in System To Accelerate Metrics Observability by 100x Us...
24mOne Pane to Rule Them All: Uniting the Prometheus Community with O...