this seems to be related to the option to reserve 12% ot the node, but the overall resource round up to 168 cpus, as there are more instance managers