Troubleshooting Python with OpenTelemetry Tracing

In the classic definition, Observability is something one step beyond monitoring; it’s how easy our system is to understand with the architecture and monitoring we have. The problem is a familiar one: we have monitoring tools but they’re not answering our question. This article shows how a Python developer can go from having traces but not answers, to fully understanding the root cause of a latency issue. The Problem: Python Monitoring without Observability Let’s lay out our scenario: we’re using a Python application with automatic monitoring from the OpenTelemetry Python SDK. It’s correctly monitoring the number of requests we’re getting, and it’s able to tell us that there are some extreme spikes in latency. This service loads a web page for users, so response times greater than 7 seconds are unacceptably slow! So, we definitely have monitoring working, in that it’s showing us there’s a problem and it’s frequent enough to be a concern for all users. Let’s dive in further and we’ll se