Operating Prometheus in a Serverless World
summary
Cloudflare builds and runs a custom Prometheus-compatible collector called the gravel gateway (implemented in Rust) that accepts standard text exposition metrics and a custom clear_mode label to encode semantics like sum, replace and family (info) so short-lived serverless functions produce correct counters, gauges and histograms. It uses time-bucketed pebbles for moving-window aggregates (means, percentiles) to expire old values, runs in production, faces scaling work at high RPS and the code is public.
AI-written summary · about
chapters (6)
- 0:00Start
- 2:37Cloudflare Invented Its Own Serverless Platform
- 9:42Push Gateway
- 10:43Aggregation Gateway
- 14:54Infometrics
- 23:55Statistical Functions
related talks
25mMonitoring in a Serverless World
03mUsing OpenTelemetry and Prometheus to monitor serverless
21mMonitoring Serverless Workloads with OpenTelemetry and Prometheus
34mFleeting Metrics: Monitoring Short-lived or Serverless Jobs with Prometheus
27mObservability with Prometheus and beyond
36mObservability with Prometheus and Beyond: Leveraging Cloud-native Technologies Outside of Cloud
33mSeeing the Forest for the Trees: An Operator’s Guide to Observability of CF with...
33mIntro: Prometheus