Unbalanced CPU load on every time

[YangYongZhi]

There are 4 nodes in my cluster, like this:
192.168.90.2 (master eligible : true, node: true)
192.168.90.3 (master eligible : true, node: true)
192.168.90.4 (master eligible : true, node: true)
192.168.90.5 (master eligible : true, node: true)

and using '_id' field as the routing field.

A index has 30,585,798 documents, and it's size is 30.67GB,

This index has 6 primary shards and 1 replica shard:
shards: 6 * 2

Many update requests will come on every day, then elasticsearch will do too many query operations, and gather it as a result list, and update them using bulk mode.

the client connects to this cluster on port 9300 as a transport node.

Problem : Only one node (192.168.90.5) load is very high every time. others had low CPU load.

Anybody can help me? thanks.

if you need more clues, I will supply them.

thanks again!

[YangYongZhi]

Elastic search version is 2.1.1

[Christian_Dahlqvist]

Are you distributing requests evenly across all nodes? How are the primary shards distributed across the nodes?

[YangYongZhi]

No, I do not do anything for distributing request.

The node has very high CPU load is 192.168.90.73, and there is only one primary shard (shard #0) on it.

Only index 'dmall_ware__ware_sku_location' have many update requests on every day.

Thanks.