Question-76: Hinted handoff repair is used while read or write?

Answer: Hinted handoff is used while writing the data, and it is the coordinator node responsibility to store the hint.  

Question-77: Can GC (Java Garbage Collection) pause make node outage?

Answer: yes, it is possible because of GC pause, your node is not responding and write request can be missed on that node. 

Question-78: Is it possible that coordinator nodes stop writing the hints for a particular node?

Answer: yes, it is possible that if a node is out for more than configured time using the parameter “max_hint_windwo_in_ms” then coordinator node will not keep the hint for that node any more.

Question-79: Coordinator node is throwing “OverloadedException” error, what does that mean?

Answer: When coordinator node find that several nodes experience brief outages simultaneously, then memory pressure buildup on the coordinator node. If the number of hints increases too much, then coordinator node refuses further writes and throws the “OverloadedException” error.

Question-80: Can you explain the read repair path in Cassandra database?

Answer: In read repair following steps are followed

• When coordinator node receives a read request, then it would look for other replicas as well on other nodes to find their digest value.
• If there is a mismatch in the data digest returned by the coordinator node and other replica. Then data would with the latest time stamp would be returned to user and other replica which were part of this process would be repaired as well.
• It is possible that a single replica does not have entire record latest. Few columns in that can be latest and few columns can be outdated. So that a new record is assembled by mixing and matching columns from different replica.
• Once the latest record is created record is written back to only the replicas involved in the request.