Keeping tabs on production
summary
Covers concrete tooling and tactics: use Elixir/OTP21 Logger with request_id and additional user_id metadata so logs correlate to a user session, note Logger’s async buffer can exhaust and flip to synchronous mode (hurts performance), and OTP21 reduces need for custom backends; forward logs to Papertrail or platform logging. Use remote console and :sys traces (cap traces at ~100), add recon for safe production tracing, use Telemetry and Telemetry.Poller for metrics, hook an Upsignal plug/logger handler for exceptions, and explore OpenTelemetry/spandex for distributed traces.
AI-written summary · about
chapters (9)
- 0:00Start
- 2:00How Do You Keep Customers Happy
- 8:50Inspecting Logs
- 24:04Recon Trace
- 26:02Production Monitoring
- 26:28Exception Tracking
- 30:32Dynamic Dispatch Library
- 32:55Telemetry Metrics
- 35:54Distributed Tracing
related talks
39mMonitoring Elixir With OpenTelemetry
33mThe Observant Developer - OpenTelemetry from code to prod
55mBeginner's guide to observability
36mIt's not in Production Unless it's Monitored
36mOperable Erlang and Elixir
32mWhat’s my App *Really* Doing in Production?
21mUsing OpenTelemetry to Build More Human-Friendly Systems
36mIt's Not in Production Unless it's Monitored