Data consistency

In a distributed file system, we keep multiple copies/replicas of file data for availability purposes. If such systems allow data mutationsA data mutation is a change in the data via random writes and appends., then data inconsistencies among replicas tend to appear for various reasons. For example, a node failure causes a mutation to fail on one of the replicas, making the data inconsistent among replicas; one replica will contain stale data while others have been updated. Clients reading data from multiple replicas will get different data, which can have consequences depending on the specific use case. Therefore, the file system that allows mutations should ensure data consistency.