HP EVA Cluster Extension Software Administrator Guide (5697-2445, July 2013)

Behavior of VM Management functionality in various disaster scenarios
This section describes various disaster scenarios and how the VM management functionality behaves
in such scenarios:
• CSV disk owning cluster node and the VM owning cluster node are in different datacenter and
the VM Role is offline:
When you try to bring the VM role online, the online operation fails because the owner nodes
of the respective resource are running in different datacenters. When a resource fails on the
cluster node, the role behaves according to the cluster role failover policies.
• CSV disk owner node and the VM owner node are in same datacenter and the VM Role is
offline:
When you try to bring the VM role online, the online operation succeeds because the owner
nodes of the respective resource are running in the same datacenter.
• The cluster node that owns the CSV disk fails, or the CSV disk fails and the CSV disk moves to
a node in the same datacenter, or the CSV disk is moved manually to a node in the same
datacenter:
In this scenario, because the CSV disk is moved to a node in the same datacenter, all the VMs
continue to run on their respective cluster nodes. There will not be any change in the ownership
of the VMs.
• The cluster node that owns the CSV disk fails, or the CSV disk fails and the CSV disk moves to
a node in the remote datacenter, or the CSV disk is moved manually to a node in the remote
datacenter:
In this scenario, all the VMs that are configured with CLX feature for VM management wait
for the CSV disk to come online on a cluster node in the remote datacenter. Because the CSV
disk is moved to a node in the remote datacenter, all the VMs that are running in the respective
cluster nodes in the primary datacenter will be live migrated by CLX to cluster nodes in the
remote datacenter. Every role has next preferred node property. Selection of the target node
for the live migration is based on how the next preferred owner property has been set for the
VM role.
Preferred Owners is not set: In this scenario, the next preferred owner node is selected randomly
among the cluster nodes present in the remote datacenter. The selected cluster node must be
up. If not, search will continue until a cluster node is found in the remote datacenter.
Preferred Owners is partially set: In this scenario, cluster nodes that are not part of the next
preferred owner node property are added to the list in the order of NodeID. Next preferred
owner that is in the same datacenter as the datacenter where CSV disk is owned is selected
from that list. The selected cluster node must be up. If not, search continues in the list until a
cluster node is found in the remote datacenter.
Preferred Owners is completely set: Because all the cluster nodes are part of the next preferred
owner node property list, the next preferred owner node is selected from this list. The selected
cluster node must be up and must be in the same datacenter as the datacenter where the CSV
disk is owned. If not, search continues in the list until a cluster node is found in the remote
datacenter.
• When VM owning node goes down and the VM moves to a node in the remote datacenter:
In this scenario, the VM fails to come online on the remote datacenter and the cluster tries to
bring it online on a node in the local datacenter.
• When VM owning node goes down and the VM moves to the next preferred node in the same
datacenter:
In this scenario, the VM tries to come online on the next preferred node in the same datacenter.
Administration 47