Streaming Recording Rules for Prometheus, Thanos, and Cortex Using...
summary
Shows how the M3 coordinator performs in-memory streaming aggregation so Prometheus/Cortex/Thanos avoid huge remote queries. A cAdvisor demo ingests metrics from tens of thousands of containers, converts a 51k-series CPU query into about 230 series and drops render time from ~20s to <1s render. Uses rollup and mapping rules, storage policies (1m/1h) and a docker‑compose dev stack with raw and aggregate Prometheus; live demo reports roughly 43x cardinality reduction and successful real‑time aggregation.
AI-written summary · about
chapters (6)
- 0:00Intro
- 3:40Prometheus Recording Rules
- 5:44M3 Streaming Aggregation
- 8:32Rollup Rules
- 15:27Demo
- 21:22Questions
related talks
19mStream vs. Batch: Leveraging M3 and Thanos for Real-Time Aggregation
26mPrometheus That Scales
30mIntro: Prometheus
20mAggregating Metrics In-Flight: Challenges, Opportunities
23mAuto-suggesting and generating recording rules for Prometheus
39mAbsorbing Thanos Infinite Powers for Multi-Cluster Telemetry
31mWhen Prometheus Can’t Take the Load Anymore
37mIntro: Cortex