I'm running Mongodb 3.2 WiredTiger on Amazon EC2 with 3 shards, each having a replica set of 3 instances. All instances have been installed identically.
EC2 instances limit IO on the amount of operations you can do per second. With this in mind, I've been surprised that our primaries and secondaries seem to write in very different ways. The primaries do roughly half the write operations of our secondaries while writing roughly 50% more in terms of bytes.
Below is a picture from our first replica set. The primary is blue, secondaries are green and orange.
Is this behavior that I can change? The secondaries are becoming problematic. The amount of data saved in terms of bytes is very small, but the secondaries are overwhelming our IO operations per second limits. The primaries have no such problems. If possible, I would want the secondaries performing writes in a way similar to the primary. Anything to reduce the WriteOps count would be helpful.