Supervisor Actor is unscheduled while managing worker Actors

Ray Core
1

How severe does this issue affect your experience of using Ray?

  • High: It blocks me to complete my task.

I’m using ray core and the Supervisor Actor Pattern, Pattern: Using a supervisor actor to manage a tree of actors — Ray 2.9.0. Although not exactly as presented in the docs.

My Supervisor ActorA creates and manages many ActorPools that each contain many ActorB. ActorA calls the various ActorPool methods (get_next, has_free, pop_idle, submit, etc.) to manage and submit work to the pools.

After a few minutes of running successfully, the Supervisor ActorA is unscheduled in favor of the many ActorBs.

I end up with many ActorBs running, although they don’t do tasks work because the Supervisor ActorA is stuck waiting to be scheduled.

Is there a way to guarantee the supervisor actor is always scheduled and running, can I assign some sort of priority that can be assigned to an instance of an Actor?

2

Just a small clarification, ActorA is displayed in the dashboard with a state of PENDING_CREATION.

3

An actor is typically never “unscheduled” by Ray. Maybe it crashed? Would you mind finding the logs for the actor’s worker and see what happened?

Also do you have a minimal repro script? Thanks.