Sometimes data replicas on one or more Cassandra nodes can get out of sync. There are a variety of reasons for this, including prior network instability, hardware failure, and nodes crashing and staying down past the three-hour hint window. When this happens, Apache Cassandra comes with a repair process that can be run to rectify these inconsistencies.
The repair process does not run on its own, and must be executed manually. Tools such as Reaper for Apache Cassandra allow for some automation, including scheduling cluster repairs.
Conditions that can cause data inconsistencies essentially do so by introducing factors that increase the entropy of the stored replicas. Therefore, Apache Cassandra employs an anti-entropy repair mechanism, ...
