Hunting for Risky Dependencies in the World of Microservices
summary
Describes a practical method that uses OpenTelemetry to record the original RPC origin and caller chain, aggregates traces into a central data lake, and programmatically compares services that actually sit on the end user path versus those meant to be isolated. Concrete results include finding 33 risky dependencies (10 internal systems and 23 client callers), catching GPay’s use of the canonical store and getting it fixed within a month, and reducing the serving surface to prevent future major outages.
AI-written summary · about
chapters (6)
- 0:00Intro
- 0:35Google Pay
- 6:31Risky Dependencies
- 9:47Finding Risky Dependencies
- 12:36Fixing Risky Dependencies
- 13:53Summary
related talks
40mMeasuring Micro-services
31mUnderstanding Microservices with Distributed Tracing
31mTroubleshoot Your RoR Microservices with Distributed Tracing
32mA Picture is Worth 1,000 Traces
58mMessage processing failed! But what's the root cause?
24mHow Deep Systems Broke Observability
32mInvestigating Performance Issues In Microservices Arch.
21mNavigating the Complexities of Service to Service Invocations: Deep and Brief Dive Into Causality