https://github.com/openshift/origin-server/pull/5482

Checked on latest devenv with the change merged, issue has been fixed.

$ ssh -t node1 service openshift-node-web-proxy restart
Stopping node-web-proxy:                                   [  OK  ]
Starting node-web-proxy:                                   [  OK  ]
Connection to node1 closed.
$ ssh node1
[root@node1 ~]# service openshift-node-web-proxy status
web-proxy (pid  20903) is running...

Can't start openshift-node-web-proxy after oo-install. However 'ssh root.15.XX service openshift-node-web-proxy restart' works.

1) oo-install failed to start service
=system_u:object_r:locale_t:s0 tclass=file
OpenShift: oo-diagnostics output - 
OpenShift: oo-diagnostics output -             Note that SELinux violations are not always logged in the audit log.
OpenShift: oo-diagnostics output - 
OpenShift: oo-diagnostics output - FAIL: test_services_enabled
OpenShift: oo-diagnostics output -       The following service(s) are not currently started:
OpenShift: oo-diagnostics output -         openshift-node-web-proxy
OpenShift: oo-diagnostics output -       These services are required for OpenShift functionality.


2) service can be started by ssh command. 
[hanli1@broker ~]$ sh test.sh
ssh root.15.89 service openshift-node-web-proxy restart
[  OK  ] node-web-proxy: [FAILED]
Starting node-web-proxy: [  OK  ]
ssh root.15.89 service openshift-node-web-proxy status
web-proxy (pid  2110) is running...

ssh root.15.89 service openshift-node-web-proxy restart
[  OK  ] node-web-proxy: [  OK  ]
Starting node-web-proxy: [  OK  ]
ssh root.15.89 service openshift-node-web-proxy status
web-proxy (pid  2172) is running...

You didn't mention how you ran oo-install (the original bug was filed against oo-install for enterprise, which is tracked in https://bugzilla.redhat.com/show_bug.cgi?id=1103138).

For origin oo-install, this change hasn't been backported into origin release 3, and would have to be tested against the nightlies (if they have even been updated since then).

For online verification, being able to restart the service over an ssh connection with a pty (ssh -t) and verifying the service status after the ssh connection that issued the restart is terminated should be sufficient.

Since for online ssh verification is sufficient, so make it as verified.

https://github.com/openshift/origin-server/pull/5520

Found another edge case.

ssh -t cloud-user.14.88 'sudo bash -c "service openshift-node-web-proxy restart; sleep 2;"'

Before this PR, service openshift-node-web-proxy status on the host after ssh is broken.

$ ssh -t ec2-54-234-181-219.compute-1.amazonaws.com 'sudo bash -c "service openshift-node-web-proxy restart; sleep 2;"'
Stopping node-web-proxy:                                   [  OK  ]
Starting node-web-proxy:                                   [  OK  ]
Connection to ec2-54-234-181-219.compute-1.amazonaws.com closed.

$ ssh ec2-54-234-181-219.compute-1.amazonaws.com "service openshift-node-web-proxy status"
web-proxy (pid  2443) is running...

Checked on devenv_4890, issue fixed.