Quildreen Motta Ribeiro
Engineer at Klarna and Illustrator
Quil is an illustrator, writer, and software engineer with a background in programming language design. They're particularly interested in making computers safer and more reliable for everyone. In their free time Quil writes horror fiction, paints with digital watercolours, and works on Crochet, a security and privacy-conscious programming language that uses types to describe and enforce security constraints.
Past Activities
Code BEAM Europe 2022
11.25 - 12.10
Know your Past to Predict your Future --- Time Travel Debugging in Production!
You've just hit a weird corner case in your production system. Looking at the code has proved hopeless. Your system's log has so much information you feel overwhelmed. How do you debug it? And how do you go about understanding all of the implications of this issue? In this talk we'll take the learnings from building specialised time-travelling debuggers for production systems and dive into how they help in these scenarios. Then we'll look into how you'd approach building one for your system, as well as the constraints that they put on your system's design.
OBJECTIVES
The goal of this presentation is to show people why they'd consider building specialised time-travelling debuggers for their production system, and look at a few approaches to building them, as well as what kind of constraints that puts in their system's design, based on debuggers the author has worked with before.
The idea of tracing itself isn't complicated, but it's not always clear what kind of trade-offs you need to make in production: you need to be more conscious of the required time and space overhead (which you can't always avoid, e.g.: when building a protocol parser). And it's not always clear how to correlate and present that data in a way that makes the information useful---most data in traces is noise, and just like looking at logs of modern production systems is useless without heavy filtering and correlation, debug traces for time-travelling need specialised analysis and visualisation tools.
AUDIENCE
This talk is aimed at engineers that run complex production systems on the BEAM, and would like better tools to understand their system's production behaviour when things go wrong.