Description of problem: This came up as part of the review for documentation MR [1]. When following the process outlined in section 16.3.2. Replacing a removed Compute node [2], the result is that the re-added compute host gets the correct index, but incorrect hostname. So for example, starting with compute-0, compute-1, and compute-2, scaling in compute-1, then readding compute-1, the node will be named compute-1 from the undercloud's POV, but the hostname (and therefore the nova-compute service) will be called compute-3. Version-Release number of selected component (if applicable): How reproducible: Steps to Reproduce: 1. As an example, start with 3 compute hosts, compute-0, compute-1, and compute-2. 2. Scale in to remove compute-1. 3. Scale out to replace the compute host. Actual results: New compute is called compute-1 from undercloud's POV, but the hostname is compute-3. Expected results: Both the node and the host should be compute-1. [1] https://gitlab.cee.redhat.com/rhci-documentation/docs-Red_Hat_Enterprise_Linux_OpenStack_Platform/-/merge_requests/9130 [2] https://jenkins.dxp.redhat.com/job/CCS/job/ccs-mr-preview/25205/artifact/titles/director-installation-and-usage/preview/index.html#proc_removing-or-replacing-a-compute-node_scaling-overcloud-nodes
The full list of steps can be seen here: http://pastebin.test.redhat.com/1052055
@Irina @jparker I guess we need to amend [1] to include Rabi's suggestion, and retest? Or perhaps start by testing, then once it's confirmed amending the procedure? [1] https://gitlab.cee.redhat.com/rhci-documentation/docs-Red_Hat_Enterprise_Linux_OpenStack_Platform/-/merge_requests/9130