Not a requirements but a best practice to have nodes clusters with identical or at least similar configuration, the reason behind this is for instance, if you have two physical servers, one with 256GB of physical RAM and another one with only 128GB, if the host with 256GB fails, the remaining host may not be able to start all virtual machines from the failed host. Not sure what you mean with that "master" configuration, but you can create DRS affinity rules to pin virtual machines to specific hosts.
---
Richardson Porto
Senior Infrastructure Specialist
LinkedIn: http://linkedin.com/in/richardsonporto