Easy anomaly Detection with PromQL
Lightningat PrometheusDay Europe 2022 · on Tuesday, 2022-05-17· at 16:24· Pavilion 3, Room F | Level 2 | Central Forum + Live Stream, Feria Valencia, Spain + Live Stream·
summary
Shows concrete PromQL recipes: use avg_over_time and stddev_over_time to compute z scores (1σ ≈70%, 2σ ≈95%) and alert when a series exceeds mean ± N·stddev, apply avg_over_time/stddev_over_time over 5m/1h/1w windows for short‑term vs seasonal baselines, create seasonal comparisons by duplicating past windows with label_replace with offset to avoid series collision, and model cross‑series expectation (building energy vs external temperature) with a correlation + error band; examples include sensor spike detection and weekly/daytime seasonality.
AI-written summary · about
chapters (7)
- 0:00Introduction
- 0:43What is an anomaly
- 1:58Anomalies in a group
- 2:49Average and standard deviation
- 3:56Seasonality
- 4:50Group
- 5:15Scenarios
related talks
30mPractical Anomaly Detection at Scale With PromQL
34mAnomaly Detection in Time Series from Scratch Using...
35mObservability’s Sixth Sense: Detecting Anomalies in Metrics
38mWatch Everything, Watch Anything: Anomaly Detection
35mObservability’s Sixth Sense: Detecting Anomalies in Metrics
40mObservability’s Sixth Sense: Detecting Anomalies in Metrics
31mFirst Day Foresight: Anomaly Detection for Observability
01h 19mIntro + Deep Dive: Prometheus