To summarize the doc change.
* existing instructions say [1]:
~~~
...
Finally, remove the node’s Compute service:
(undercloud) $ source ~/stack/overcloudrc
(overcloud) $ openstack compute service list
(overcloud) $ openstack compute service delete [service-id]
~~~
This needs a note which says that in case the compute is being re-deployed with the exact same name, those steps are not needed.
The service can either be disable with a comment to know why it is down or leave it as is (detected as down) until the node is re-provisioned. When the service gets disabled it needs to be re-enabled after a node with the same hostname is up again.
e.g.
$ openstack compute service set --disable --disable-reason "gets re-provisioned" compute-1.localdomain nova-compute
$ openstack compute service list --long
...
| 80 | nova-compute | compute-1.localdomain | nova | disabled | up | 2018-07-13T14:35:04.000000 | gets re-provisioned |
+----+------------------+--------------------------+----------+----------+-------+----------------------------+---------------------+
Same for the neutron agent:
$ openstack network agent set --disable --description "gets re-provisioned" 3134ae01-f3d2-432c-886c-dcf4b4ac2a3c
When up again enable the services:
$ openstack compute service set --enable compute-1.localdomain nova-compute
$ openstack network agent set --enable --description "gets re-provisioned" 3134ae01-f3d2-432c-886c-dcf4b4ac2a3c
[1] https://access.redhat.com/documentation/en-us/red_hat_openstack_platform/12/html-single/director_installation_and_usage/#sect-Removing_Compute_Nodes