MongoDB secondary records are twice as large as primary

Question

I am running Mongodb 3.2 WiredTiger on Amazon EC2 with 3 shards, each of which has a replica set of 3 instances. All instances are installed the same way.

EC2 instances limit IOs to the number of operations you can perform per second. With that in mind, I was surprised that our primaries and minor editions seem to be written in a completely different way. Primaries do about half the write operations of our secondary, with about 50% more bytes.

Below is a picture from our first set of replicas. The primary is blue, the secondary are green and orange.

Can I change this behavior? Secondary ones become problematic. The amount of data saved in bytes is very small, but secondary users suppress our I / O operations per second. Primaries have no such problems. If possible, I would like secondary records to be performed in the same way as primary ones. Anything that could reduce the WriteOps score would be helpful.