Description of problem: On 4.12 cluster, SSH over NodePort does not work and it looks a regression issue. 1. on 4.12, ssh using virtctl works $ virtctl -n default ssh centos7-r1tv26qtediogn3y fix_vm_overview The authenticity of host 'vmi/centos7-r1tv26qtediogn3y.default:22 (10.0.206.71:6443)' can't be established. ECDSA key fingerprint is SHA256:jfNOw1xVDixwSHrSF8pMFlWjQE6AqURUqjXlKgrohAQ. Are you sure you want to continue connecting (yes/no)? yes 2. on 4.12, SSH over NodePort does not work, it timeouts after wait for enough time $ ssh centos.uit-412-1208.cnv-qe.rhcloud.com -p 32603 ssh: connect to host console-openshift-console.apps.uit-412-1208.cnv-qe.rhcloud.com port 32603: Connection timed out 3. on 4.10.z, SSH over NodePort is working $ ssh fedora.uit-410-1207.cnv-qe.rhcloud.com -p 30085 fix_vm_overview The authenticity of host '[console-openshift-console.apps.uit-410-1207.cnv-qe.rhcloud.com]:30085 ([10.0.155.152]:30085)' can't be established. ED25519 key fingerprint is SHA256:xAuDrudzQ6qK/wNXOcFTcVU+FPFts5+nC8zaFG3OtZI. This key is not known by any other names Are you sure you want to continue connecting (yes/no/[fingerprint])? yes 4. on 4.11, SSH over NodePort is working ssh centos.uit-411-1212.cnv-qe.rhcloud.com -p 31370 fix_vm_overview The authenticity of host '[console-openshift-console.apps.uit-411-1212.cnv-qe.rhcloud.com]:31370 ([10.0.206.118]:31370)' can't be established. ED25519 key fingerprint is SHA256:p9nj8lF/YGftKBph0i7CxjyuuCIAiPTGq4S9gtoS9cA. This key is not known by any other names Are you sure you want to continue connecting (yes/no/[fingerprint])? yes Version-Release number of selected component (if applicable): v4.12.0 How reproducible: 100% Steps to Reproduce: 1. create a VM and start it 2. enable the ssh service over nodeport in vm details tab 3. copy the ssh command and try it in command line Actual results: SSH does not work and timeouts after wait for enough time Expected results: SSH is working Additional info:
From 4.11 to 4.13 the service command is unchanged and also the service template used. Basically, we create the same service across all versions, and what change is just the NodePort that the cluster decides to use. Does the cluster configuration is changed between versions? @gouyang
I retested the problem on bare metal 4.12 cluster, the issue is not happening. Close the bug for now.
*** This bug has been marked as a duplicate of bug 2165895 ***