The active thread always in one dataNode，other dataNodes have very few active thread or no active thread

Elastic Stack Elasticsearch
7

1.the command is 'GET /_cat/thread_pool/search?v&h=port,ip,id,name,active,queue,rejected,completed'

2.I had changed my nodes ,but the result is just like this

9300 10.17.107.38 IMEntercTkK0XD8Giked7g search 8 0 0 38763
9300 10.17.107.56 2byzeSwjQSat9pBlAbe0DA search 7 0 0 39184
9300 10.17.107.36 ynuFvHGSTVmlb58FpYO9ig search 32 1 0 72920
9300 10.17.107.37 MsEnlGbxQIK5gkJeQPcr4Q search 8 0 0 117160
9300 10.17.107.39 e_mR6vgcTLq3oI00KQtkcQ search 8 0 0 69293

as you can see , 10.17.107.36 has many 'active' ,even has 'thread',and it has high cpu