Configuring Disaster Recovery with Metro

Learn about Metro Availability, which provides an automated disaster recovery plan.

Requirements

• Access to the OVHcloud Control Panel
• Access to your clusters via Prism Central
• You need to have three Nutanix clusters within the OVHcloud infrastructure with Pro or Ultimate packs if you have an OVHcloud packaged service on both clusters in the P.R.A. These three clusters will need to be at remote sites for maximum security.
• You must have less than 5 ms of latency between the two replicated clusters. Please note that latency is not covered by SLAs.

Introduction

We will set up a two-way disaster recovery plan between two clusters with this hardware:

• A Nutanix cluster in Datacenter 1, with virtual machines replicated in Datacenter 2.
• A Nutanix cluster in Datacenter 2, with virtual machines replicated in Datacenter 1.
• A Nutanix cluster in Datacenter 3, with Prism Central to serve as a witness in the disaster recovery plan.

We will only use one vRack, which will contain:

• The three Nutanix clusters.
• Load balancers.
• Additional IP addresses on the rtvRack.

Below is the diagram showing the three sites:

Instructions

Step 1 Configuration

     Step 1.1 Interconnection of the three clusters

     Step 1.2 Delete the Prism Central records for the Datacenter 1 and Datacenter 2 clusters

     Step 1.3 Register both clusters on Prism Central in Datacenter 3

     Step 1.4 Adding IP Addresses for iSCSI Connections on All Three Clusters

     Step 1.5 Creating two Storage Containers

     Step 1.6 Move virtual machines to the Storage Container

     Step 1.7 Creation of a category to be used when implementing the P.R.A.

     Step 1.8 Add virtual machines in categories

     Step 1.9 Setting up synchronous replications between Datacenter 1 and Datacenter 2

     Step 1.10 Create Subnets for Disaster Recovery Plan

     Step 1.11 Implementation of disaster recovery plans

Step 2 Validate Disaster Recovery Plan

     Step 2.1 Monitoring the Disaster Recovery Plan

     Step 2.2 Live migration of virtual machines from Datacenter 1 to Datacenter 2

     Step 2.3 Operations after a migration

     Step 2.4 Execute the Disaster Recovery Plan in Real Condition

We will implement this disaster recovery plan step by step.

The cluster configuration information used in our guide is as follows:

• Datacenter 1 cluster:

  • Server 1: VM address CVM 192.168.0.21, hypervisor IP address AHV 192.168.0.1.
  • Server 2: VM address CVM 192.168.0.22, hypervisor IP address AHV 192.168.0.2.
  • Server 3: VM address CVM 192.168.0.23, hypervisor IP address AHV 192.168.0.3.
  • Prism Element virtual address: 192.168.0.100.
  • Prism Element iSCSI address: 192.168.0.102.
  • Prism Central IP address: 192.168.0.101.
  • Gateway: 192.168.3.254.
  • Mask: 255.255.252.0
  • Cluster version: 6.5

• Datacenter 2 cluster:

  • Server 1: VM address CVM 192.168.1.21, hypervisor IP address AHV 192.168.1.1.
  • Server 2: VM address CVM 192.168.1.22, hypervisor IP address AHV 192.168.1.2.
  • Server 3: VM address CVM 192.168.1.23, hypervisor IP address AHV 192.168.1.3.
  • Prism Element virtual address: 192.168.1.100.
  • Prism Element iSCSI address: 192.168.1.102.
  • Prism Central IP address: 192.168.1.101.
  • Gateway: 192.168.3.254.
  • Mask: 255.255.252.0
  • Cluster version: 6.5.

• Datacenter 3 Cluster:

  • Server 1: VM address CVM 192.168.2.21, hypervisor IP address AHV 192.168.2.1.
  • Server 2: VM address CVM 192.168.2.22, hypervisor IP address AHV 192.168.2.2.
  • Server 3: VM address CVM 192.168.2.23, hypervisor IP address AHV 192.168.2.3.
  • Prism Element virtual address: 192.168.2.101.
  • Prism Element iSCSI address: 192.168.2.102.
  • Prism Central IP address: 192.168.2.100.
  • Gateway: 192.168.3.254.
  • Mask: 255.255.252.0
  • Cluster version: 6.5.

In addition to this guide, you can use these documents:

• Nutanix hyperconvergence
• advanced tools

Step 1 - Configuration

Step 1.1 - Interconnection of the three clusters

The first step is to interconnect the three clusters on the same OVHcloud vRack.

Use this guide to connect your clusters: Interconnect clusters through the vRack. To connect the three clusters, use the instructions provided in the guide:

• Datacenter 1 clusters in the vRack dedicated to Datacenter 3.
• Datacenter 3 clusters in the vRack dedicated to Datacenter 2.

When you have finished configuring your vRack, you will have these elements in your vRack:

• 9 dedicated servers (3 per cluster)
• 3 public IP addresses
• 3 Load Balancers

The three clusters are currently accessible from the Prism Central URL of each cluster.

Step 1.2 - Delete the Prism Central records for the Datacenter 1 and Datacenter 2 clusters

To implement a disaster recovery plan solution with Metro Availability, a cluster witness is required to automate tasks in the event of one of the clusters becoming unavailable. The cluster witness is located on a Prism Central virtual machine.

The Datacenter 3 cluster will host the Prism Central virtual machine for the three clusters, and serve as a cluster witness for the disaster recovery plan between Datacenter  1 and Datacenter 2.

Disabling Prism Central on the Datacenter 1 cluster

Connect via SSH to the Prism Element cluster in Datacenter 1 :

ssh nutanix@private_ip_address_prism_element_Roubaix
Enter Prism Element password

Run this command to remove Prism Element from the Prism Central configuration:

ncli multicluster remove-from-multicluster external-ip-address-or-svm-ips=private_ip_address_central_roubaix\
username=admin password=pwd_pe_Roubaix force=true

This message appears when disconnecting from Prism Central.

Cluster unregistration is currently in progress. This operation may take a while.

Enter this command:

ncli cluster info

Disconnect from Prism Element and connect via SSH on the Prism Central virtual machine in Datacenter 1.

ssh nutanix@private_ip_address_prism_central_roubaix
Enter Prism Central password

Enter this command:

python /home/nutanix/bin/unregistration_cleanup.py cluster_uuid_prism_element_Roubaix

Disabling Prism Central on the Datacenter 2 cluster

Log in to the Prism Element cluster in Datacenter 2 via SSH.

ssh nutanix@private_ip_address_prism_element_Gravelines
Enter Prism Element password

Enter this command:

ncli multicluster remove-from-multicluster external-ip-address-or-svm-ips=private_ip_address_prism_central_Gravelines\
username=admin password=pwd_pe_Gravelines force=true

This message appears when disconnecting from Prism Central.

Cluster unregistration is currently in progress. This operation may take a while.

Enter this command:

ncli cluster info

Note the value of Cluster UID that should be in this form xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx

Disconnect from Prism Element and connect via SSH on the Prism Central virtual machine in Datacenter 2.

ssh nutanix@private_ip_address_prism_central_Gravelines
enter Prism Central password
python /home/nutanix/bin/unregistration_cleanup.py cluster_uuid_prism_element_Gravelines

Step 1.3 - Registration of the two clusters on the Prism Central in Datacenter 3

Log in to the Prism Element in Datacenter 1 via SSH:

ssh nutanix@private_ip_address_prism_element_Roubaix
enter Prism Element password

Run this command:

ncli multicluster register-to-prism-central username=admin password=passwod_admin\ external-ip-address-or-svm-ips=private_ip_address_prism_central_Erith

This message appears:

Cluster registration is currently in progress. This operation may take a while.

Wait and enter this command:

ncli multicluster get-cluster-state

If the cluster is connected to Prism Central in Datacenter 3, you will see this information:

Registered Cluster Count: 1

    Cluster Id                : xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx
    Cluster Name              : Prism-Central-Erith-FQDN
    Is Multicluster           : true
    Controller VM IP Addre... : [private_ip_address_prism_central_Erith]
    External or Masqueradi... :
    Cluster FQDN              :
    Controller VM NAT IP A... :
    Marked for Removal        : false
    Remote Connection Exists  : true

Log in to Prism Element in Datacenter 2 via SSH:

ssh nutanix@adresse_ip_prism_element_Gravelines
Enter Prism Element password from Gravelines

Run this command:

ncli multicluster register-to-prism-central username=admin password=passwod_admin_Erith external-ip-address-or-svm-ips=private_ip_address_central_Erith

This message appears:

Cluster registration is currently in progress. This operation may take a while.

Wait and enter this command:

ncli multicluster get-cluster-state

If the cluster is connected to Prism Central in Datacenter 3, you will see this information:

Registered Cluster Count: 1

    Cluster Id                : xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx
    Cluster Name              : Prism-Central-Erith-FQDN
    Is Multicluster           : true
    Controller VM IP Addre... : [private_ip_address_prism_central_Erith]
    External or Masqueradi... :
    Cluster FQDN              :
    Controller VM NAT IP A... :
    Marked for Removal        : false
    Remote Connection Exists  : true

From a web browser, log in to the URL from Prism-Central to Datacenter 3, you will see the three clusters.

The Prism Central virtual machines in Datacenter 2 and Datacenter 1 are no longer being used. You can stop them.

In the main menu, click VMs in the Compute & Storage submenu.

Select the Prism Central virtual machines in Datacenter 2 and Datacenter 1 and click Guest Shutdown from the Actions menu.

Step 1.4 - Verifying IP Addresses for iSCSI Connections on All Three Clusters 

From the Prism Central dashboard, click the link to the Datacenter 3 cluster.

On the Prism Element dashboard, click the cluster name in the top left-hand corner.

Scroll down the window, verify the IP address under ISCSI Data Services IP, and click Save. 

From the Prism Central dashboard, click the link to the "Datacenter 2 cluster".

On the Prism Element dashboard, click "the cluster name" in the top left-hand corner.

Scroll down the window, verify the IP address under ISCSI Data Services IP, and click Save. 

From the Prism Central dashboard, click on the link to the "Datacenter 1 cluster".

On the Prism Element dashboard, click the "cluster name" in the top left-hand corner.

Scroll down the window, verify the IP address under ISCSI Data Services IP, and click Save. 

Step 1.5 - Creating Two Storage Containers

We will create two Storage Containers with the same name, one in Datacenter 1 and the other in Datacenter 2.

From the Prism Element main menu, click Storage Containers in the Compute & Storage submenu.

Click Create Storage Container.

Type UsedForDR in Name, choose the Datacenter 1 cluster in Cluster, and click Create.

Click Create Storage Container.

Type UsedForDR in Name, choose the Datacenter 2 cluster in Cluster, and click Create.

In the list of Storage Containers, you will see two Storage Containers with the same name. One on the Datacenter 1 cluster and the other on the Datacenter 2 cluster.

Step 1.6 - Moving virtual machines to the Storage Container

We will move the virtual machine storage to the Storage Container we have created.

Connect via SSH on the Prism Element of the Datacenter 1 cluster:

ssh nutanix@private_ip_address_Prism_element_Datacenter1
Enter the Nutanix account password of Prism Element

Run this command for each VM we will move to the Storage Container, replacing vmname with the name of the virtual machine (in our disaster recovery plan, we have two virtual machines in Datacenter1, one on Windows and one on Linux).

acli vm.update_container vmname container=UsedForDR
Enter the Nutanix account password of Prism Element

Log in to the Prism Element of the Datacenter 2 cluster via SSH:

ssh nutanix@private_ip_address_Prism_element_Gravelines
Enter the Nutanix account password of Prism Element

Execute this command for each VM that we will move to the Storage Container, replacing vmname with the name of the virtual machine (in our disaster recovery plan, we have three virtual machines in Datacenter 2, one on Windows, another on Linux, and the gateway that gives access to the Internet).

acli vm.update_container vmname container=UsedForDR
Enter the Nutanix account password of Prism Element

Step 1.7 - Creation of a category to be used when implementing the P.R.A.

We will create a category with two values in Prism Central to assign the virtual machines involved in replication.

Scroll through the main menu, click Categories in the Administration submenu.

Click New Category.

Type Protected VM in Name, add the <DATACENTER 1> and <DATACENTER 2> values in Values, and click on the next button Save.

The category appears in the list and is ready to use.

Step 1.8 - Adding virtual machines in categories

We will assign two virtual machines on the Datacenter 1 cluster in one category and three virtual machines on the Datacenter 2 cluster in another category.

From the Prism Central main menu, click VMs in the Compute & Storage submenu.

Select the two virtual machines in Datacenter 1 on the left, then on the Actions menu, and click Manage Categories.

Add the category ProtectedVM: <DATACENTER 1>, then click Save.

Select the three virtual machines in Datacenter 2 on the left, then Actions, and click Manage Categories.

Add the category ProtectedVM: <DATACENTER 2>, then click Save.

Step 1.9 - Setting up synchronous replications between Datacenter 1 and Datacenter 2

Synchronous replication allows permanent replication with 0 seconds of data loss.

Replication setup between Datacenter 1 and Datacenter 2

On the Prism Central main menu, click Protection Policies in the Data Protection submenu.

Click Create Protection Policy.

Type <DATACENTER 1>-TO-<DATACENTER 2> in Policy name, keep Local AZ, and click Select Cluster in Primary Location.

Choose the Datacenter 1 cluster and click Save.

In the Disaster Recovery message, click Enable.

The system checks that everything is correct before enabling Disaster Recovery.

Once the checks are complete, click Enable to enable the Disaster Recovery option.

Click Enable again.

Your Disaster Recovery option is being activated.

Keep Local AZ, select the cluster in Recovery Location, and click Save.

Click + Add Schedule.

Choose Synchronous for Protection Type and Automatic for Failure Detection Mode. Then click Save Schedule.

Click Next.

Select the category ProtectedVM : <DATACENTER 1> and click Add.

Click Create.

Virtual machines in Datacenter 1 are now replicated to Datacenter 2. You must wait for the first full replication to have permanent replication.

Replication setup between Datacenter 2 and Datacenter 1

Replication can be two-way. We will now create a replication from Datacenter 2 to Datacenter 1.

Click Create Protection Policy.

Choose as name <DATACENTER 2>-TO-<DATACENTER 1> in Policy Name, keep Local AZ, and choose Datacenter 2 cluster in Primary Location. Then click Save.

Keep Local AZ, select the Datacenter 1 cluster, and click Save.

Click + Add Schedule.

Choose Synchronous for Protection Type and Automatic for Failure Detection Mode. Then click Save Schedule.

Click Next.

Select the category ProtectedVM: <DATACENTER 2> and click Add.

Click Create.

A second protection strategy is in place.

Step 1.10 - Create Subnets for Disaster Recovery Plan

We will create subnets that will be used to test disaster recovery plans.

For each existing subnet, a test network is required. On the two clusters of the Disaster Recovery Plan, we have three production subnets.

• based on VLAN 0.
• infrastructure on VLAN 1.
• production on VLAN 2.

We will therefore create three additional subnets on the Datacenter 2 and Datacenter 1 clusters with these names:

• testing on VLAN 100.
• testinfra on VLAN 101.
• production on VLAN 102.

Use this guide to create VLANs on your Nutanix clusters: isolate production management machines.

In the Prism Central Subnets dashboard, you will see six new subnets.

Step 1.11 - Implementation of disaster recovery plans

Now that the replications and subnets are in place, we will implement automated or manual disaster recovery plans on demand to:

• migrate virtual machines on the fly between the two clusters
• Test that replication is working properly
• Automatically restart the VMs that are members of the P.R.A. in the event of a failure of one of two clusters.

Creation of a disaster recovery plan for the Datacenter 1 cluster

In the main menu of Prism Central, click Recovery Plans in the Data Protection submenu.

Click on Enable Disaster Recovery on the left.

Normally, the recovery plan must be activated as indicated with the message Disaster Recovery enabled. Click on the right to close this window.

Click Create New Recovery Plan.

Choose this information:

• Recovery Plan Name: Recovery VM from <DATACENTER 1> to <DATACENTER 2>.

• Primary Location: Local AZ.

• Primary Cluster: cluster in Datacenter 1.

• Recovery Location: Local AZ.

• Recovery Cluster: cluster in Datacenter 2.

• Failure Execution Mode: Automatic.

• Execute failover after disconnectivity of: 30 seconds.

Then click Next.

Click + Add VM(s).

Select both virtual machines and click Add.

Click Next.

Click OK. Got it.

Click Stretch networks.

Click Proceed.

Choose the VLANs that will be used during the IP like this:

• Primary

  • Production production
  • Test Failback: testproduction

• Recovery

  • Production: production
  • Test Failback: testproduction

Then click Done.

Creation of a disaster recovery plan for the Datacenter 1 cluster

The Disaster Recovery Plan has been created for the Datacenter 1 site. Click Create Recovery Plan to create the Datacenter 2 Disaster Recovery Plan.

Choose this information:

• Recovery Plan Name: Recovery VM from <DATACENTER 2> to <DATACENTER 1>.

• Primary Location: Local AZ.

• Primary Cluster: cluster in Datacenter 2.

• Recovery Location: Local AZ.

• Recovery Cluster: cluster in Datacenter 1.

• Failure Execution Mode: Automatic.

• Execute failover after disconnectivity of: 30 seconds.

Then click Next.

Click + Add VM(s).

Select the three virtual machines and click Add.

Click Next.

Click Stretch networks.

Click Proceed.

Choose this information:

• Primary

  • Production: basis
  • Failback test: test

• Recovery

  • Production: basis
  • Failback test: test

Then click + Add Network Mapping.

Choose this information:

• Primary

  • Production infra
  • Test Failback: testinfra

• Recovery

  • Production: infra
  • Test Failback: testinfra

Then click + Add Network Mapping.

Choose this information:

• Primary

  • Production: production
  • Test Failback: testproduction

• Recovery

  • Production: production
  • Test Failback: testproduction

Then click Done.

Both disaster recovery plans are in production.

Step 2 - Validate Disaster Recovery Plan

Step 2.1 - Monitoring the Disaster Recovery Plan

Using the Commit to Disaster Recovery Plan Option

You can validate the disaster recovery plan via Prism Central.

Click on the Recovery VM from Datacenter 1 to validate and test.

Click Validate.

Select the Datacenter 1 cluster for Entity Failing Over From and the Datacenter 2 cluster for Entity Failing Over To. Then click Proceed.

The recovery plan has been validated. Click Close.

Test Disaster Recovery Plan

We can test the disaster recovery plan without impacting production. The test creates virtual machines with different names on the destination cluster in the VLANs created earlier.

Click Test.

Select the Datacenter 1 cluster for Entity Failing Over From and the Datacenter 2 cluster for Entity Failing Over To. Then click Test.

Click Execute Anyway.

Go to the VM dashboard in Prism Central and you will see the test virtual machines that are created with the replicated data.

Return to your recovery plan and click Clean-up test entities to remove the test virtual machines.

Click Clean Up.

Step 2.2 - Live migration of Datacenter 1 virtual machines in Datacenter 2

On a fully operational infrastructure, it is possible to move virtual machines from one cluster to another without any service downtime.

Go to a virtual machine in Datacenter 1 that is part of the recovery plan. We will ping the OVHcloud DNS server 213.186.33.99.

Return to your recovery plan and open the More menu, then click Failover.

Choose Planned Failover, then select the Live Migrate Entities box.

Select the Datacenter 1 cluster for Entity Failing Over From and the Datacenter 2 cluster for Entity Failing Over To.

Then click Failover.

Type Failover and click Failover.

Hot migration is in progress.

The migration was completed successfully without any service downtime.

You can go back to the virtual machine and see that the ping continues to work even if the virtual machine has been moved from one cluster to another.

Step 2.3 - Operations after a hot migration

After a migration, it is necessary to reverse the replication and operation of the disaster recovery plan.

Reverse Replication

On the Prism Central main menu, click Protections Policies in the Data Protection submenu.

Click on the protection plan named <DATACENTER 1> TO <DATACENTER 2>.

Click Update.

Position the mouse below the Datacenter 1 cluster name in Primary Location and click Edit.

Check the <DATACENTER 2> cluster instead of the Datacenter 1 cluster.

Click Save.

Click Update Location.

Position the mouse below the Datacenter 2 cluster name in Recovery Location and click Edit.

Select the <DATACENTER 1> cluster instead of the Datacenter 2 cluster.

Click Save.

Click Update Location.

Click Next.

Click Update.

Replication is reversed, click the button to close the protection plan.

Disaster Recovery Plan Reverse

In the main menu of Prism Central, click Recovery Plans in the Data Protection submenu.

Click Recovery VM from <DATACENTER 1> to <DATACENTER 2>.

On the More menu, click Update.

In locations, put the Datacenter 2 cluster in Primary Clusters and the Datacenter 1 cluster in Recovery Clusters and then click Next.

Click Proceed.

Click Next.

Choose this information:

• Primary

  • Production: production
  • Test Failback: testproduction

• Recovery

  • Production: production
  • Test Failback: testproduction

Click Done.

Step 2.4 - Execute the Disaster Recovery Plan in Real Condition

We will simulate a total loss of connection to Datacenter 2 where three virtual machines are located in the disaster recovery plan (the Internet gateway and two other virtual machines).

Log in to the command line and ping the public address of the gateway.

## Ping from a remote linux console
ping xx.xx.xx.xx
Reply from xx.xx.xx.xx: bytes=32 time=21ms TTL=58
Reply from xx.xx.xx.xx: bytes=32 time=21ms TTL=58
Reply from xx.xx.xx.xx: bytes=32 time=23ms TTL=58
Reply from xx.xx.xx.xx: bytes=32 time=20ms TTL=58

Leave the ping command to run continuously and return to Prism Central.

In the main menu, click VMs in the Compute & Storage submenu.

The three virtual machines in the disaster recovery plan are functional.

All three nodes in the Datacenter 2 cluster will be disconnected.

Return to the console that is pinging to the gateway, and you will see a connection loss.

Reply from xx.xx.xx.xx: bytes=32 time=20ms TTL=58
Reply from xx.xx.xx.xx: bytes=32 time=21ms TTL=58
Reply from xx.xx.xx.xx: bytes=32 time=20ms TTL=58
Reply from xx.xx.xx.xx: bytes=32 time=20ms TTL=58
Reply from xx.xx.xx.xx: bytes=32 time=20ms TTL=58
Request timed out.
Request timed out.
Request timed out.
Request timed out.

In Prism Central, click the top right on the tasks to display the task launch, including Recovery plan execute.

Go back to the text console and you will see that the ping works again.

Request timed out.
Reply from xx.xx.xx.xx: bytes=32 time=20ms TTL=58
Reply from xx.xx.xx.xx: bytes=32 time=19ms TTL=58
Reply from xx.xx.xx.xx: bytes=32 time=18ms TTL=58
Reply from xx.xx.xx.xx: bytes=32 time=18ms TTL=58
Reply from xx.xx.xx.xx: bytes=32 time=19ms TTL=58
Reply from xx.xx.xx.xx: bytes=32 time=19ms TTL=58

Go to Prism Central in the virtual machine management, you will see the three virtual machines of the recovery plan in duplicate. They are marked as started, but in reality, only the ones restarted in Datacenter 1 are working.

We will reconnect the three nodes in the vRack to return to normal mode.

After the recovery, the virtual machines on the original cluster are still visible but are turned off. You can delete or keep them if problems occur on the VMs that are being rebooted.

You can view the history of Disaster Recovery actions in Prism Central.

Click the button in the top right-hand corner to go to the Prism Central configuration.

Left-click Witness and click View Usage History.

The list of events appears, click Close to close.

Go further

Browse all the Nutanix guides.

Browse the rest of the OVHcloud US documentation for guides on our other products and services.

Learn more about Nutanix on OVHcloud on the OVHcloud US website.