Intro: Prometheus
Talkat KubeCon + CloudNativeCon North America 2019 · on Wednesday, 2019-11-20· at 16:25· Pacific Ballroom, Salon 20-22 - Marriott Marquis San Diego Marina Hotel·
summary
Prometheus is a single Go binary that scrapes instrumented apps and exporters (Go/Python/Java/Ruby/C/Rust), stores samples in a local TSDB with a write‑ahead log and compaction to reduce SSD wear, and uses remote_write to Thanos/Cortex for long‑term storage. PromQL does ratios, histograms and 99th‑percentile latency; alert rules feed an Alertmanager. Typical production claims: >1M samples/sec ingest and examples of 10 million series on one host. HA runs two servers scraping the same targets and statefulsets or PVs preserve data.
AI-written summary · about
chapters (33)
- 0:00Introduction
- 0:27What is Prometheus
- 1:47Limitations of Prometheus
- 2:29History of Prometheus
- 3:03Prometheus Architecture
- 4:13Prometheus Metrics
- 4:54Prometheus Targets
- 5:05Prometheus Service Discovery
- 5:51Prometheus Ecosystem
- 7:18Major Selling Points
- 8:30Prometheus Example Time Series
- 9:00Prometheus Querying
- 11:22Prometheus Histogram
- 11:57Prometheus Adhoc
- 12:16Prometheus Graph
- 12:28Persistent dashboards
- 12:41Prometheus alerting
- 13:49Operational simplicity
- 14:21Efficiency
- 14:52Remote Storage
- 15:26Service Discovery
- 15:45Disco Service Discovery
- 16:29Conclusion
- 17:10Questions
- 17:48Write amplification
- 19:09Write rate
- 20:06Avoid duplicate values
- 20:46Compression
- 26:41Prometheus Federation
- 27:43Global Views
- 29:04Authentication TLS
- 29:44Prometheus Server
- 31:53One Alternative Approach
