The db.serverStatus().opLatencies section is added to mongos, giving latencies of reads, writes, commands, and transactions from mongos, i.e. including the time spent between mongos and mongod. mongos opLatencies thus include the full latency an operation spends inside all of MongoDB, whereas in a sharded cluster the mongod opLatencies do not include the time spent going from mongos to mongod or back from mongod to mongos.

Description of Linked Ticket

Similar to mongoD, mongoS should capture latency metrics. This will make it possible to inspect the average latency trend.