Description of problem: One of our VIP customers struggled to deploy overcloud with pCPUs from NovaComputeCpuSharedSet list being isolated from OS processes: CPU affinity of containerized composable services seem to be global, so he failed to do this using tuned. When I tried to understand if there is a workaround from TripleO perspective, I have found out that paunch supports setting various memory limits and defining CPU sets for specific containers [1]. cpuset_cpus setting should be translated via composable service's configuration, but currently such option is only implemented for nova-libvirt container [2] (there is no such options for different containers/composable services). I am reporting this RFE to ask TripleO team to implement some THT variable that limits cpuset for ALL containers when is is set. Setting severity to High to ensure that this would be addressed in RHOSP 17 as this feature is quite important to customer. [1] https://github.com/openstack/paunch/blob/stable/train/paunch/builder/podman.py [2] https://github.com/openstack/tripleo-heat-templates/blob/stable/train/deployment/nova/nova-libvirt-container-puppet.yaml#L259 [3] https://access.redhat.com/documentation/en-us/red_hat_openstack_platform/16.2/html-single/advanced_overcloud_customization/index#sect-Using_Customized_Overcloud_Heat_Templates