How To Rename Metrics Without Impacting Somebody’s Observabili...
Talkat KubeCon + CloudNativeCon Europe 2025 · on Thursday, 2025-04-03· at 15:00· Level 1 | Hall Entrance N10 | Room G·
summary
Prototype implements on-the-fly metric versioning for Prometheus by attaching an OpenTelemetry-style schema URL label and applying autogenerated transformation files at query time; it converts names, labels and even units (histogram ms→s) without changing producers or doubling storage. Implementation is about 1000 lines as a layer between PromQL and storage, uses the weaver tool for schema validation/codegen, and demo shows fast, correct translations and gradual rollouts with pinned schema versions so dashboards and alerts keep working.
AI-written summary · about
chapters (8)
- 0:00Metric renaming challenges
- 3:32Metric renames are problematic
- 9:30Strategies to mitigate rename issues
- 15:50Automating versioned metric references
- 20:45Implementing seamless metric renames
- 25:22Live demo: schema-based renaming
- 28:35Supporting metric conventions
- 30:03Key learnings and next steps
related talks
28mObservability Project Updates
31mEvolution of Metric Monitoring and Alerting: Upgrade Your Prometheus Today
24mLeveraging OpenTelemetry for Your Prometheus Pipeline
47mGet Instrumented: How Prometheus Can Unify Your Metrics
13mConnecting Prometheus and OpenTelemetry Data for Faster Troublesho...
32mOpenTelemetry Semantic Conventions and How to Avoid Broken Observability
28mOpenTelemetry developments in Prometheus over the last year
05mLightning Talk: Schema Inference