Inside the cluster: replication and storage
Modules 2 and 3 treated the cluster as a reliable box that stores records and hands them out. This module opens the box. It shows what happens when a broker stops or crashes, how many controllers you can lose, and how records leave the disk again: by age, on request, or by being replaced with a newer value for the same key.
By the end of this module you can: - explain how followers, the ISR and the high watermark work, and predict what clients notice when a leader stops or crashes; - restart brokers safely, rebalance leaders, and recognise a cluster that has lost its controller majority; - set retention so data is really gone by a deadline, and spot consumers that will fall behind it; - delete records on request, and choose a strategy for erasing one person's data; - use compacted topics and tombstones, and rebuild a service's state from them.
Lessons
- Leaders, followers and failover
- Segments, retention and deleting data
- Compacted topics: the latest value per key
Then practise with the exercises and check yourself with the module quiz.