Question-46: If a node goes down for some time because of outage and comes back, then how it gets the missed writes during the outage?

Answer: If a node comes back after outage then repair mechanism is used to get the data which are missed for this node. The repair mechanism can be manual which can be initiated using the command “nodetool repair” or there is a mechanism exists known as hinted handoff. It's depend on how long was the outage, and accordingly you would use repair mechanism.

Question-47: In Cassandra how does it determine on which node the data is stored?

Answer: Using the primary key it determines on which node the data is stored.

Questions-48: The first replica of the data is considered master replica?

Answer: No, every replica of a data record in the Cassandra cluster is equally important and there is no master replica concept.

Question-49: What do you mean by consistent hashing?

Answer: When a hashing key is derived from the partition key, it should be always the same and consistent. Whenever a new node is added or removed in cluster then data distribution across cluster should have minimum reorganization of the data. 

Question-50: What are the virtual nodes or vnodes in the Cassandra cluster?

Answer: Suppose you have a four-node grouped together (where each node represents physical machine) then you can say it is a 4-node cluster. But now assume that if each physical node in cluster represent 10 virtual nodes (this is still not a virtual machine; it is just a conceptual division based on some configuration). Then in the cluster there are in total 40 virtual nodes. This virtual node helps the data distribution with the finer granularity and it is considered better than single token architecture. In this cluster actually having four physical nodes but it has 40 virtual nodes. And tokens are distributed across 40 virtual nodes which gives many benefits, specially reducing the load on one particular physical node, by distributing the load across remaining nodes in the cluster.