Session
Break It Before Production Does: Chaos-Testing a Raft Database Cluster
Building a distributed database is hard. Proving it survives real-world failures is harder. In this talk we open the ArcadeDB test lab and show how we break our own Raft-based high-availability cluster on purpose, on every change and every week.
You'll see how we use Testcontainers and Toxiproxy to spin up real multi-node clusters and inject the failures production loves: leader crashes, rolling restarts, network partitions, split brain, latency and packet loss. Then we go further: chaos tests that run unattended for one to several hours, hammering the cluster with load while faults keep coming, and verifying that every replica converges to consistent data.
Distributed bugs are notoriously hard to explain, so we built animated diagrams of every scenario. You'll watch leader elections, log replication and partition recovery step by step, exactly as the tests run them.
For Java developers and architects who build or run distributed systems and want to sleep better at night.
Roberto Franchini
Principal Software Engineer @ ArcadeData
Turin, Italy
Links
Please note that Sessionize is not responsible for the accuracy or validity of the data provided by speakers. If you suspect this profile to be fake or spam, please let us know.
Jump to top