Going to move from mgo to this Go driver. Having a sharded cluster with replica sets for each shard.

What I did with mgo (and in C++ implementations as well): Connecting directly to every single replica set member without mongos to have better control of my connections and load of each node.

I'm wondering if this is a good approach or there is a better best practise. Thought about connecting to shards and let the driver handle the replica sets. But I will lose some of the control with this.