Question-71: What is hinted handoff repair?

Answer: Let's assume because of some reason node does not able to receive a particular write request which is coming from a coordinator node. Then in this case the coordinator nodes responsibility to preserve the data which needs to be written on that node and that would be saved as hint. Hence, whenever nodes come back online, the coordinator hands off that hint to that node. This node can catch up with the required pending or missed writes. 

Question-72: What is the read repair?

Answer: During the read path a query, storage engine assemble data from several nodes, and it is the coordinator nodes responsibility to get all the replicas from the other nodes wherever they are stored in cluster and compare those replicas. If it's found any replica outdated then coordinator node sends the most recent version of replica to the outdated node. However, keep in mind it does not have to collect all the replicas, it only collects enough replica to satisfy the replication factor and repair applies only on the node that participated in that read operation. Hence, if all replicas are not read then it is still possible that some replicas are outdated.

Question-73: What is Anti-entropy repair?

Answer: if you want to repair every replica then use the nodetool repair utility, it compares the data across all the replica and then update the data to the most recent version. And for regular maintenance you should use this utility. Simply if you want to update every replica with the most recent record then use the Anti-entropy repair.

Question-74: What are the basic properties of NodeSync repair utility?

Answer: NodeSync utility runs in the background and does the following things

• It is continuously validating the data across all the replica.
• It is always running, and has a very low impact on cluster performance.
• This does not require any manual intervention.
• It completely replaces anti-entropy repairs. 

Question-75: What are the segments when node sync utility is used?

Answer: When node sync is enabled then it uses local data ranges for the repair. And this local data ranges again splitted into small segments, and behave as validation save point.