On top of a single SCVM failure scenario, if multiple SCVM failures cause all data replicas of a virtual machine to become unavailable, the corresponding virtual machine will also become unavailable. If the disks are physically intact, the data replicas can be restored after the SCVMs recover, and the affected virtual machines will return to normal operation.
If more than half of the SCVMs running ZooKeeper and Meta services in the cluster fail, the metadata services will become unavailable, and the cluster will be unable to operate normally.
