A Picture is Worth 1,000 Traces
summary
Claims that aggregate distributed-trace analysis and targeted visual diffs find failures faster than single-trace hunting. Uses Jaeger with synthetic, city-specific black-box monitors that produce golden-trace baselines, compares bad traces to the model, and links diffs into full spans for SQL, logs and error messages. Notes Uber-scale context: north of 3000 microservices, integration with OpenCensus/OpenTelemetry collectors, tail‑based sampling, and a reported reduction in ops time from 30 minutes to 2 minutes to isolate responsible services.
AI-written summary · about
chapters (12)
- 0:00Intro
- 3:54Agenda
- 6:37Why of distributed tracing
- 11:14What went wrong
- 11:52Context Correlation
- 12:48Distributed Tracing
- 15:02Summary
- 19:21Challenges
- 19:44The Approach
- 22:06Outage Detection
- 22:38Debugging
- 26:23Conclusion
related talks
51mNavigating Distributed Traces
34mReduce Complexity with Distributing Tracing
34mWould You Like Some Tracing With Your Monitoring?
35mTracing is More Than Traces: The Insights in Trace Aggregates
07mWhy Tracing is a Lifesaver for E2E Tests of Distributed Systems?
24mDistributed Tracing : New DevOps Foundation.
49mDistributed Tracing: How the Pros Debug Concurrent and Distributed Systems
14mTurning Synthetic Traces into Gold: Monitoring for User Journeys