We’ve been subjecting the YugaByte DB to various performance and load tests over the last few months. While these load tests provide a baseline for how the system performs, and subjects it to common node-failure scenarios, it would be useful to see how YugaByte holds up in the presence of more adversarial failures, such as network partitions, clock drift, etc.
Jepsen is a widely used framework to evaluate databases’ behavior under different failure scenarios. It allows for a database to be run across multiple nodes, and create artificial failure scenarios, as well as verify the correctness of the system under these scenarios.
We have begun work to test YugaByte as well under the Jepsen framework. We will be subjecting
the nodes to various failures and evaluating the responses, in order to validate correctness. Stay tuned for further updates regarding these experiments, and the findings.