Discussed this with Sean Mooney and targeting for 16.1.2
The general process to document is:
1. You upgrade the overcloud all the way up to running "openstack overcloud upgrade converge"
2. You correct your VMs thread policy (either by migration or resizing)
3. You convert from vcpu_pin_set to cpu_dedicated_set occurs
In addition, Sean said:
yes although there are more subtelties i.e. you should not convert form vcpu_pin_set to cpu_dedicated_set
if vms are on a host unless the value set for vcpu_pin_set will be reused exactly the same for cpu_dedicated_set
if you are going to adjust which cpus are defiend in the set for any reason the host should not have vms running on it.
so that shoudl be called out in the doc but at a high level the ordering you have is correct.
we expect new deployment of 16.1 to start with cpu_dedicated_set but for upgrades they should
continue to use vcpu_pin_set untill the upgrade is fully complete and only chage the config option afterwards.
The plan is to add a new section on updating compute pinning functionality here:
https://access.redhat.com/documentation/en-us/red_hat_openstack_platform/16.1/html-single/framework_for_upgrades_13_to_16.1/index#concluding_the_overcloud_upgrade
Will work with Sean to document the main process and any additional nuances.