Recently we tried implementing the sharding concept in MongoDB. . For this we configured 3 VMs (each 50GB) - 1 VM as config server, 1 VM as router, shard, and DB server, and the other VM also as router, shard, and DB server.

When we are inserting the data, from single spark RDD to this sharding, data population to the table is slow, and bulk write exception is coming. But when we did this manually using a simple java program we can populate more data. And the same documents are going to both the shards.

Any help on this is highly appreciated. MongDB version - 3.2.10