This is how my architecture looks like for testing / bench-marking purpose.

1. Config server running with single node replica set (on x86 build)
2. Single mongos (on x86 build)
3. 4 Shards running with standalone mongods (on different build)

I've imported 10M documents and enabled hashed Sharding. With this 10M total data size reaches up-to ~1.3 GB.  As I've applied hashed Sharding, it should balance chunk well and should start migrating the chunks over to the different shards.  But after waiting over longer period of time, it hasn't migrate single chunk over. 

When I've converted standalone mongods to single node replica set on all of the shards, it started migrating chunks and after some time, it has precisely balanced the chunks across all shards. 

My question is: 

1. was running standalone mongod on each shard was an issue?
2. Running sharded cluster components in different environments  (mongos and config on one build and mongod in another build) would be causing this issue?

Thanks!
Viraj