Hello,I am deploy ray cluster manually,and sometimes I borrow machines from other teams and add them to my ray cluster to run some heavy tasks.
The task running on these machines are expensive, and I wish them finished success,then I stop the node graceful.
But while some tasks finished,new task will scheduled to the node continuely,So it will never be a moment no task running on the node.
In k8s,we can do this by taint the node,so no more pod will scheduled to the node,but I can’t any way to archive this in ray.So is there any way to do this?