Kafka Mistakes - You Won't Work Long Enough to Make All of Them

Explore common Kafka mistakes and their solutions in this 26-minute conference talk from Devoxx Poland 2019. Dive into the pathology of code, examining hotspots, commit issues, replication traffic, and CPU performance. Learn about client crashes, exception handling, and timeout exceptions. Discover monitoring techniques and cleanup policies. Gain valuable insights to avoid pitfalls and optimize your Kafka implementation through real-world examples and practical tips.

Intro
Pathology of code
I'm going to operate!
Hotspots - why
Hotspots - diagnostics
Commit hotspots
Offset commits
Replication traffic
CPU IO wait
Avg idle
Quick tip
Specimen 1: recap
Client crushes and bad exception handling
Good, bad or ugly?
Know your exceptions
Hazmat suits on!
Too much TimeoutExceptions?
What happened?
Precautions
What can go wrong?
Monitoring on the cheap • You don't need Confluent Platform to have monitoring
Specimen 4
Cleanup policies
Specimen 5
Hint 2