OpenTelemetry 101
summary
Adopts OpenTelemetry across an Erlang service fleet, wires traces into Jaeger/Zipkin and metrics into Prometheus/Grafana via an external OpenTelemetry Collector so per-language exporters disappear, and forks an Erlang client for testing. Implements spans, attributes and baggage for request context, and uses bound instruments (counters and value recorders) for per-processor frame counts and sizes, pushing async GPU and sync frame metrics; prototypes run with Docker compose and successfully surface traces and metrics.
AI-written summary · about
chapters (20)
- 0:00Start
- 8:01Open Telemetry
- 9:34Benefits of Using an Open Standard
- 10:55What Is Tracing
- 13:38Open Telemetry Tracing Api
- 15:12Spans
- 15:58Attributes
- 20:40The Implementation
- 29:17Resource Usage
- 31:19Open Telemetry Collector
- 34:25Summary
- 36:40Metrics
- 38:19Meters
- 38:29Definitions of Open Telemetry
- 41:28Instrument Type
- 41:46Counter
- 43:37Value Recorder
- 46:29Bound Instruments
- 49:58Reduce Cardinality
- 53:25Standard Patterns
related talks
35mOpenTelemetry for Busy People
39mMonitoring Elixir With OpenTelemetry
30mOpenTelemetry 101
38mPractical introduction to OpenTelemetry tracing
22mOpenTelemetry – An Observability Framework for Cloud-Native Software
46mThe State of OpenTelemetry
32mOpenTelemetry feat. logs, metrics and traces
01h 08mOpenTelemetry: The First Release, What’s Next, and How to Get Involved