etcdctl cluster-health shows each node is health while one node is not working

[dearboll]

Hi. I find a problem.When I build my etcd cluster and want to verify whether everything is ok. I use etcdctl member list command and it shows as below:
1f31b23c41b42e33: name=infra2 peerURLs=http://IP_2:2380 clientURLs=http://127.0.0.1:2379,http://IP_2:2379 isLeader=false
6f7bfee130f3c5e6: name=infra0 peerURLs=http://IP_0:2380 clientURLs=http://127.0.0.1:2379,http://IP_0:2379 isLeader=true
b05fda0c29c5a1d2: name=infra1 peerURLs=http://IP_1:2380 clientURLs=http://127.0.0.1:2379,http://IP_1:2379 isLeader=false
And when I run etcdctl cluster-health and it shows as below:
member 1f31b23c41b42e33 is healthy: got healthy result from http://127.0.0.1:2379
member 6f7bfee130f3c5e6 is healthy: got healthy result from http://127.0.0.1:2379
member b05fda0c29c5a1d2 is healthy: got healthy result from http://127.0.0.1:2379
But in fact, I do not run etcd on the other two nodes, so I think the health check is just on 127.0.0.1:2379, how can I use it on the IP_1/IP_2/IP_3? I think it may be related with clientURLs=http://127.0.0.1:2379,http://IP_2:2379, health check is just on the first URL. And how can I change the order of it?
My --listen-client-urls flag is written as http://172.31.55.27:2379,http://127.0.0.1:2379