Kafka Topics: Pitfalls and Insights

Keeping track of the size of a Kafka topic and watching it vary over time helped us uncover unexpected patterns that might seem mysterious to the inexperienced cluster administrator. Even with limits set, the reported size can vary vastly. Periodic spikes upon deletion of data can be seen, with a frequency that varies from topic to topic. Deleting data from Kafka topics can be even more disconcerting, as under certain conditions information long past the retention policy can be unburied from the depths of the topic.