Jepsen and ScyllaDB: Putting Consistency to the Test

ScyllaDB has undertaken rigorous testing to ensure data consistency in its databases, utilizing both in-house tools and the renowned Jepsen testing framework, led by Kyle "Aphyr" Kingsbury. Jepsen, an industry standard since 2013, provides a comprehensive approach to testing distributed systems and uncovered some unexpected consistency issues in ScyllaDB's lightweight transactions (LWT), prompting improvements in documentation and operational practices. Despite initial confidence in their LWT implementation, ScyllaDB discovered and resolved several bugs, including issues in data checksums and list operations. The testing process highlighted the need for stronger coordination during topology changes, leading ScyllaDB to plan a transition to Raft-based consistency models. This commitment to quality and the collaboration with Jepsen has reinforced ScyllaDB's dedication to refining its product and ensuring robust consistency across its features, with ongoing tests to prevent regressions in future releases. The upcoming ScyllaDB Summit will offer further insights into these developments, featuring presentations by Kyle Kingsbury and Konstantin Osipov.