Prometheus Deep Dive
summary
Panel answers focus on concrete ops choices: Prometheus intentionally avoids built-in clustering and long‑term storage and instead uses the remote API and third‑party systems like Cortex or a Thanos sidecar that uploads TSDB blocks to object storage, adds a query layer and downsampling (CloudFlare testing at scale). The in‑memory head block holds ~two hours of samples and memory scales roughly 8 KB per active series. Migration 1.x→2.x tooling exists but is slow; recommended pattern is one Prometheus per cluster and federate or attach long‑term storage.
AI-written summary · about
chapters (12)
- 0:00Introduction to Prometheus
- 1:35Project philosophy
- 4:05Long-term storage options
- 9:22Maintainer backgrounds
- 12:24Alerting and configuration
- 15:50PromCon conference
- 18:05Scaling and multi-cluster
- 23:49Migration and upgrades
- 25:06Thanos integration
- 27:04Anomaly detection and ML
- 30:27Community governance
- 34:44Ecosystem and integrations
related talks
01h 19mIntro + Deep Dive: Prometheus
32mPrometheus Intro and Deep Dive
36mPrometheus - Intro, Deep Dive, And Open Q+A
36mDeep Dive: Prometheus
37mJourney inside Prometheus: Architecture, updates, outlook
42mPrometheus Intro and Deep Dive
23mPrometheus – What it is, What is New, and What is Coming
30mIntro: Prometheus