The Great Cardinality Disasters of Our Time
Talkat KubeCon + CloudNativeCon North America 2019 · on Wednesday, 2019-11-20· at 17:20· Room 17AB - San Diego Convention Center, Mezzanine Level·
summary
Prometheus deployments hit real limits: queries over 100,000 series slow or time out, a single server struggles past ~10 million series, and mislabels can generate 300,000 alerts or write 80,000 samples every 15s. Long error strings explode the TSDB symbol table (one case from 5GB to 60GB), and mtail misconfiguration follows 50,000 log files until relabeling reduces that to 79. Use TSDB analyze/promtool, relabeling, admin delete+compaction, and soft limits to find and remove the offending series.
AI-written summary · about
chapters (9)
- 0:00Introduction to cardinality
- 2:36Understanding Prometheus scale
- 5:23Alerting rule disaster
- 9:50Label value bloat
- 14:53Histogram configuration
- 20:34Metric name mismanagement
- 24:27Diagnostic techniques
- 28:30Solving label anomalies
- 32:08Q&A and mitigation strategies
related talks
20mThe Sins of High Cardinality
56mUnwiring High Cardinality
29mBeyond TSDB: Unlocking Prometheus with Parquet for Modern Sc...
11mAggregating, Alerting and Graphing on Millions of Prometheus T...
01h 19mIntro + Deep Dive: Prometheus
39mHow To Take Prometheus Planet Scale: Massively Large Scale Metrics Installations
33mPrometheus Project Intro
23mPrometheus – What it is, What is New, and What is Coming
