I would like to have the following situation:
- a ray cluster running several nodes
- an application gets started on node A
- some actors for that application can run on any node in the cluster (no problem here)
- some actors need to run on the SAME node where the application got started: is there a way to ensure this?
More generally, at the time an actor gets started, my code knows if it should be restricted to the same node as where it gets started or not. How can this be done?
Rationale: some actors do not need external resources or only resources which can get shared between nodes in some way. However there are also actors which need resources only present on the node where they get started.