Automated, Distributed Systems Testing for Kubernetes Controllers

Explore automated testing techniques for Kubernetes controllers in this conference talk. Delve into common bug patterns and learn about Sieve, an innovative tool designed to systematically test and fortify controllers against various distributed system challenges. Discover how Sieve has uncovered and resolved critical safety issues in popular Kubernetes controllers for applications like Zookeeper, Cassandra, RabbitMQ, MongoDB, and XtraDB. Gain insights into the importance of controller correctness, the impact of asynchrony, unexpected failures, networking issues, and controller restarts on system stability. Understand the potential consequences of safety violations, such as incorrect deletion of StatefulSets and PVCs. Follow along as the speakers discuss the tool's internals, analyzer functionality, and its application in improving observability and addressing other bug patterns. Conclude with a comprehensive understanding of how automated testing can enhance the reliability and safety of Kubernetes controller implementations.

Introduction
Kubernetes Controllers
Controllers
Challenges
Problems
Solution
correctness
general purpose checks
demo
observability gaps
other patterns
internals
analyzer
conclusion