Hi My intention is to statically partition the slaves in my cluster such that jobs belonging to a specific role has resources carved for them, and the rest of the resources in the slave are for the remaining jobs.
This is what my understanding is to enable this: On mesos-slave: I start mesos-slave with resources carved out for role1 as: --resources="cpus(role1):1;mem(role1):2048;disk(role1):32768" On scheduler (I am using aurora), I have the FrameworkInfo protobuf have the role set to 'role1' instead of the default value of '*'. On mesos-master, I start with the following flag: --roles=role1 With this, I see the offer from the slaves only contain the cpu, mem and disk as mentioned for 'role1'. How do I enable the slave to offer the complete set of resources (not just for 'role1') but resource allocation for jobs be based on whether the role is 'role1' (use "cpus(role1):1;mem(role1):2048;disk(role1):32768") whereas for all other jobs, it picks up from the rest of the resources within the slaves of the cluster. Or is it supposed to work differently? Any pointers would be great. Thanks Anindya
