A data storage system includes controller nodes and a storage component. The
controller
nodes couple the storage component to a host. The storage component includes a
controller and sleds with disk drives. The host and the storage component form
an arbitrated loop. When the arbitrated loop is down, the controller removes the
storage component from the arbitrated loop so the controller, the sleds, and the
disk drives form an internal loop. When the internal loop is also down, the controller
tests each of the sleds individually and marks those that are not responding properly.
If a sled is responding properly the controller tests each of the drives in that
sled individually and marks those that are not responding properly. The controller
adds back into the internal loop the sleds and the drives that are not marked as
bad, and then the storage component back into the arbitrated loop.