+ cancellable=True, retry_wait=1, max_retry_wait=180):
+ # If a ShutdownActor is cancellable, it will ask the
+ # ComputeNodeMonitorActor if it's still eligible before taking each
+ # action, and stop the shutdown process if the node is no longer
+ # eligible. Normal shutdowns based on job demand should be
+ # cancellable; shutdowns based on node misbehavior should not.