Ruby Monitoring State of the Union
summary
Claims that instrumentation is unit testing for operations and shows concrete primitives: counter meter histogram timer with examples of queue counts, request rates, request-size distributions and timing percentiles. Advocates collecting metrics in a double-buffered in-process registry so reporters can pull without adding runtime dependencies, using exponential weighted moving averages for rates and reservoir/priority sampling for percentiles. Recommends separating reporters (Graphite, statsd, Librato) into independent gems so instrumentation stays dependency‑free and testable.
AI-written summary · about
chapters (16)
- 0:00Start
- 2:37Modern Infrastructure
- 4:01Cont. Deployment
- 10:47Anti-Pattern
- 11:21Current Status
- 14:16Devs Require
- 15:18Ideal Instrumentation
- 18:07Timer
- 19:51Registry
- 20:40Reporters
- 22:26Challenges
- 22:54Counter
- 23:12Meter
- 24:30Histogram
- 26:59Metriks
- 27:43Future Work
related talks
36mIt's not in Production Unless it's Monitored
36mMetrics - What's your code actually doing?
36mIt's Not in Production Unless it's Monitored
28mMonitoring Rails apps using Open Source
53m.NET Rocks Live
27mGrand Treatise of Modern Instrumentation and Orchestration
32mWhat’s my App *Really* Doing in Production?
30mRails Performance monitoring: A primer for developers