The CPU useage of elasticsearch three nodes is not balance

[zekaifeng (kaifeng ze)]

elasticsearch version：7.10.2
nodes：3
indice config：3 primary shard & 1 Replica

I use {"query": {"match_all": {}} to query 10 pieces of data, and the CPU useage of the three nodes is not balance.
node-0:

node-1:

node-2:

[Christian_Dahlqvist]

What is the full output of the cluster stats API?

How are you running the test? How many concurrent connections are you using? Have you verified that the connections and the load are evenly spread out across the nodes? Is there any other load on the cluster at the time you test the queries?

Is this imbalance causing any issues? As far as I can see there seems to be plenty of resources available.

[zekaifeng (kaifeng ze)]

Using esrally to test, I have tried to connect directly to the node and the domain name, all of which are the same.
The cluster stops all other business operations during query.

[Christian_Dahlqvist]

What about the other questions?

Have you checked the number of Rally connections to the different nodes?

I do not see this being something to worry about.

I also wonder what you are looking to getcfrom this test, which does not seem very realistic.