Does the Scheduler support instance affinity?
No the Scheduler supports Service `affinity but not instance affinity. Let’s assume a cluster has 5 nodes. Service A consists of nodes 1 and 2 and Service B consists of nodes 3, 4 and 5. You can specify that a job class has service affinity for Service B. This means that the jobs belonging to that job class will ONLY run on nodes 3, 4 or 5. At the job level you can specify Instance Stickiness. This implies that the Scheduler will attempt to run a job on the instance that it last ran on. However if this is not possible (instance is either down or overloaded) the job will run on another instance.
