A replica-set secondary maintains a replicated copy of the primary ' s dataset. The primary records data-changing operations in its oplog, and secondary members continuously copy and apply those oplog entries asynchronously to keep their data synchronized. This redundancy allows an eligible secondary to participate in an election and potentially become primary if the current primary becomes unavailable. mongos, option B, is the query-routing process used in a sharded cluster; it does not store an application dataset copy. An arbiter, option C, participates in replica-set voting and elections but deliberately does not maintain a copy of the database data. A shard, option D, is a data partition in a sharded cluster and is normally implemented as a replica set in production, but “shard” is not the replica-set member role that replicates the primary ' s data. Secondary behavior is central to MongoDB high availability, read-scaling options, backup strategies, and failover. Administrators monitor secondary replication lag and health to ensure the copies remain sufficiently current. Therefore the member responsible for maintaining copies of the primary ' s data is the secondary, making option A correct.

Study Guide reference/topic:Cluster Reliability and Data Resilience - replica-set primaries, secondaries, oplog replication, and redundancy.

===========