1. start kafka cluster with 3 nodes 2. create a topic with replication factor = 3, min.in.sync.replicas = 2 3. start a producer 4. start a kafka.js consumer 5. find and kill the broker that is acting as a topic leader 6. wait for a rebalance, check that there are still 2 active brokers with a new leader 7. kill the new leader, now the cluster has only 1 node, which is not enough to commit any new data 8. get one of the dead nodes back up, wait for it to sync up 9. observe the producer to recover 10. observe kafka.js consumer to recover, but to fetch new records at 30s intervals