Your App is not a Black Box
summary
Travis runs 25,000 open source projects and does about 10,000 test runs daily, so visibility replaces guessing: logs are one‑line lograge entries for web, verbose contextual logs for workers (five VMs per worker), exceptions go to Haystack, metrics come from instrumenting code with ActiveSupport Notifications and flushing via metricsd or statsd to Librato or Graphite, and a custom event monitor (Ember + Redis pubsub) visualizes request journeys across 10 deployable apps and four environments.
AI-written summary · about
chapters (8)
- 0:00Introduction and background
- 4:53Understanding app internals
- 12:08Logging strategies
- 17:41Exception monitoring
- 20:16New Relic for non-web apps
- 22:28Metrics and visualization
- 33:17Building internal tools
- 37:48Conclusion and Q&A
related talks
42mIt's not how good your app is, it's how good you want it to be
21mMonitoring Your Apps is not About Devops
32mWhat’s my App *Really* Doing in Production?
52mIs your JVM app flying blind? Unmask hidden issues with Observability superpowers!
41mScalable Observability for Rails Applications
44mMillions of Apps: What we've learned
39mRunning Ruby Apps
38mMillions of Apps: What we've Learned