Poor performance upgrading from 1.4.2 to Elasticsearch 2.3.4

We are trying to upgrade from 1.4.2 to 2.3.4.

Both clusters are sized the same
3 masters
15 data nodes (c3.8xlarge) 64GB 32cores 2 * 320GB SSD 10Gbit ethernet

Our load average on the current cluster is anywhere from 10-15 and when we tried 2.3.4 it went up to 40.

We have 4 indexes
24GB 15 shards 1 replica
1GB 15 shards 2 replica
1GB 15 shards 2 replica
83GFB 25 shards 2 replica

Query

Profile

Hot Threads