When there are existing hugepages mounts on compute hosts, and active osp flavors that are using hugepages for active VMs, we need to be able to adopt those VMs without being shutdown during edpm-ansible bootstrap / configure_os tasks that configure hugepages. As the linked issue shows, unexpected changes to the mounted hugepages devices instantly kill all workloads.

Edpm-ansible does not apply mounts for hugepages devices, when its edpm_default_mounts and edpm_extra_mounts vars are empty.
 
We must document that starting condition for node sets created during adoption, and ensure that it will always be met in our testing as well.
 
We also need another automated tests, where we try to adopt a system that uses huge pages in a running VM instance started before adoption, without reconfiguring anything for hugepages after adoption is done. (or only reconfiguring EDPM that does not cause workload issues)
 
And another test for applying hugepages re-configuration post-adoption, with a delayed reboot enabled. This test should check that mounting hugepages devices specified in edpm_extra_mounts and edpm_default_mounts  does not impact running workloads. If such a live-mounting is not supported and requires live-migrating instances off the node, before doing anything to the hugepages mounts, we should delay it, like we do for reboot perhaps.