Journey inside Prometheus: Architecture, updates, outlook
summary
Covers Prometheus internals: it uses an HTTP pull model (with Pushgateway for ephemeral jobs), stores metrics in a single-process TSDB, and uses PromQL for queries. Warns that label cardinality drives memory and cost, can monitor tens of thousands of objects and handle up to ~100k datapoints/sec on powerful machines, and has no native sharding/clustering so production requires external systems like Thanos, Cortex or Mimir for long-term storage and scalability.
AI-written summary · about
chapters (12)
- 0:00Prometheus monitoring
- 2:11Monitoring tools history
- 6:09Prometheus origin
- 9:08Prometheus as TSDB
- 13:50Metric cardinality
- 16:20Prometheus architecture
- 20:30Metric integration
- 22:36Querying with PromQL
- 25:46Scalability limits
- 28:40Long-term storage
- 30:25OpenTelemetry standard
- 35:21Q&A
