That should be possible, but will have performance impacts / additional configurations and potential misbehavior. But in general, it should work for Yarn, but not for MRv1.
Please let me know if it’s feasible to have hadoop cluster with data nodes running on multiple Operating systems. For instance few data nodes running on windows server and others on linux based OS (RHEL,centOS).
If above scenario is feasible then please provide configuration settings required in various xml files(hdfs-site.xml,core-site.xml,mapred-site.xml,yarn-site.xml) and environment files(hadoop-env.sh/hadoop-cmd.sh) for windows and linux data nodes and namenode.