There are three ways to stop workload from running on a scavenged host:
A cluster administrator closes a host. Any running workload is allowed to run until it has finished. New workload cannot be assigned. Slots become unavailable for allocation by EGO.
A cluster administrator closes a host and the resource is reclaimed. Any running workload is terminated after a grace period. New workload cannot be assigned. Slots become unavailable for allocation by EGO.
The scavenging agent closes the host after a predefined threshold is reached. If fastrelease is enabled, any running workload is terminated without a grace period. If fastrelease is not enabled, any running workload is terminated after a grace period. By default, fastrelease is not enabled. Once the scavenging agent closes the host, new workload cannot be assigned. Slots become unavailable for allocation by EGO.