Paweł graduated from the AGH University of Science and Technology in Kraków, Poland, and followed his studies with a doctorate in the field of natural language processing. He started programming in Erlang in 2007, and worked for Klarna AB (Stockholm, Sweden) for a few years, developing fintech solutions. In 2012 he joined Erlang Solutions Ltd., where he’s had the opportunity to design, implement and optimize highly scalable Erlang systems, finally becoming the lead of the MongooseIM messaging platform team.
After years of building custom tracing tools to debug complex Erlang systems, in 2017 I decided to make them reusable and public as erlang_doctor, followed by its Elixir counterpart, ex_doctor. Debugging issues in complex Erlang systems such as MongooseIM was a real-world battlefield for me, where I refined the tools over the years with the help of suggestions from my colleagues at Erlang Solutions. Their current versions allow extensive tracing with minimal impact on the running system, and querying the collected data afterwards.
In this talk, I will show real-life examples of diagnosing elusive and subtle issues. Such experiences allowed me not only to keep improving the tools, but also to make my own code easier to debug, and - in a broader sense - easier to reason about. Are there any coding practices that could make you get rid of a bug in a matter of minutes rather than days? Or maybe there are some anti-patterns to avoid? And how do tracing, instrumentation, logging and metrics complement one another? These questions are not easy to answer, but I will share practical insights that can help you improve your own Erlang system.
Key Takeaways:
Target Audience: