Nutanix offers a “Metro Availability” option that is a policy applied on a datastore, which effectively spans the datastore across two sites. This is accomplished by pairing a container on the local cluster with one on a remote site and then synchronously replicating data between the local (active) and remote (standby) containers. When metro availability is enabled, everything in the active container is replicated synchronously to the remote container.
Figure: The example of Metro Availability
Site A
Nutanix Cluster A (3 nodes).
Protection Domain 1 – Active Controller.
Protection Domain 2 – Standby Controller.
The protected VMs enabled with “Should DRS Affinity Rule” on Controller 1.
Site B
Nutanix Cluster B (3 nodes).
Protection Domain 1 – Standby Controller.
Protection Domain 2 – Active Controller.
The protected VMs enabled with “Should DRS Affinity Rule” on Controller 2.
Site C
Witness virtual server.
Planned Maintenance
1. Update the DRS Affinity Rules in Site A.
2. Move the VMs to Site B by vMotion.
3. Promote the Metro Protection Domain (PD) in Site B.
4. Disable the PD in Site A.
5. Re-enable the PD in Site B.
Unplanned Failure
1. Site B detects outage of Site A and acquires witness lock (20 seconds).
2. Protection Domain automatically becomes active (Primary) on Site B.
3. vSphere HA will restart all VMs in Site B.
4. No administrator operations necessary.
Unplanned Failback
1. Update the DRS Affinity Rules in Site B.
2. Move the VMs to Site A by vMotion.
3. Promote the PD in Site A.
4. Disable the PD in Site B.
5. Re-enable the PD in Site A.
Example of Failure Scenario 1
The connection loss between Site A & Site B, the virtual machines continue to run.
Example of Failure Scenario 2
The connection loss between Site A & Site B, and between Site A & Site C. The virtual machines on Site A continue to run.
Example of Failure Scenario 3
Site B outage or complete network failure. The virtual machines on Site A continue to run.
Victor Wu
Chief Architect, Blogger, Author at Dell EMC Knowledge Sharing & Packt
Hi! Nice work. Is it possible to get the same solution with Nutanix Acropolis?
LikeLike
Sorry, Metro availability is only supported on ESXi hypervisor.
LikeLike
For scenario 2 I would have thought Site A would be isolated hence IO shutdown? Therefore VMs on site B would remain running.
LikeLike
Sure, the VMs on site B are remain running when all nodes shutdown on site A.
LikeLike
In case of Failure scenario 3, shouldn’t the VMs on Site B restart on Site A?
LikeLike
Yes. The VMs on Site B will restart on Site A.
LikeLike