Distributed Tracing in Python

02:15 PM - 02:40 PM on July 17, 2016, Room CR7

Yuri Shkuro

Audience level:
novice

Description

We take a look at distributed tracing as an irreplaceable tool for troubleshooting performance issues in highly distributed systems built with microservices.

Abstract

The rise of microservices and cloud-based architectures brought a new area in scaling both the distributed systems and the teams developing them. But by making the systems highly distributed, we also made them hard to understand and analyze. Debugging and finding performance bottlenecks became more complex, and the tools and techniques that worked well in the past do not do such a good job anymore.

In this talk we present the mechanics of distributed tracing systems that are designed to address these problems. We will dive into how Python microservices can be instrumented, often with a mere configuration change, to provide rich performance data to the tracing system. We will show real world examples of how tracing has helped in troubleshooting latency issues.