I have two kubernetes environments. One is v1.10.4 and the other is v1.13.5. When I deploy a exactly same service yaml file into these two environments and using curl command to check the nodeport, I cannot get response from 1.13.5 version environment. The service yaml file exposes nodeport:31218 on both environment
Under 1.10 version, I can use curl command to check the service nodeport and get response.
root@k8s-v110:~# curl -XGET localhost:31218/consumers/admin/api_key
{"total":1,"data":[{"created_at":1557977118000,"main_key":true,"enabled":true,"ref_role":1,"id":"6ec7c50e02ce","expired_time":1573529118000,"key":"5c9e9b3c5066","consumer_id":"b9eb389a3c59"}]}Under 1.13 version, If I use localhost for curl command, I did not get any response until I press ctrl+c
k8s-v113 [~] [root] # curl -XGET localhost:31218/consumers/admin/api_key
k8s-v113 [~] [root] # curl -XGET 10.111.33.10:31218/consumers/admin/api_key
{"total":1,"data":[{"created_at":1557906877000,"main_key":true,"enabled":true,"ref_role":1,"id":"68b629a3faeb","expired_time":1573458877000,"key":"60eee61045bc","consumer_id":"e0f2d6bf66ce"}]}I just wondering is there any enhancement on k8s. Thus, I cannot use localhost directly?
There isn't any change on the NodePort. However, those 2 clusters might be configured differently, specifically the one with version 1.13 such that kube-proxy doesn't bind to the localhost (127.0.0.1) but only specific network interfaces. Since it is the component responsible of NodePort. You can check --nodeport-addresses parameter on the kube-proxy to see which ips it binds to. If it is not specified or 0.0.0.0, it binds to all interfaces, and problem would be at somewhere else.
Also check if localhost is resolved correctly to 127.0.0.1 via nslookup localhost, and also try curl'ing 127.0.0.1 instead:
curl -XGET 127.0.0.1:31218/consumers/admin/api_key