Cluster-Set

The goal is to realize the value of WSFC's Cluster-Set with ECX.

Implementing Cluster-Set by EC (planning and investigation)

Gary's premise 2022.08.02

• Ignore network address difference across migration.
• LiveMigration still needs WSFC.
• Firewall going to be configured
• Hyper-V Replica
  • Router VM will not be required
• 2 node clusters
• MD resource will not be used
• Scripts will manage intra-cluster VM moves and replication changes between clusters
• Manually move the VM Cross-Cluster using Hyper-V Replica's "Planned Failover" option:
  • Stop the VM on a node in cluster-A
  • Select option to reverse the direction of replication
  • Select option to start the VM on a node in cluster-B
• LM requirement : optional or mandatory
• WSFC requirement : optional or mandatory

Yoshida's Ideas 2022.08.03

• The values of Cluster-Set
  • Scalability
  • Load Balancing
  • Can be managed fully by CLI
  • Fault domains and Availability set
• Implement LB on Cluster-Set
• Implement the function for deletion of (script, MD, SD) resource from clp.conf. clpcfset does not have the function.

WSFC with Hyper-V Replica notes 2022.09.14

Relocation of a VM from one server rack to another is possible without using Cluster-Set

Testing Configuration

• Two WSFC clusters of 2 nodes each were created
  *Note that clusters can have more nodes and be in the same domain or on different subnets.
• Hyper-V Replica Broker was added as a role to each WSFC cluster to facilitate cross-cluster VM migration
• Hyper-V Replica server was configured on each cluster from the Hyper-V Replica Broker role in Failover Cluster Manager.
  • By configuring this on both clusters, the replication direction can be reversed when a VM is moved from one cluster to another.
  • A specific server in a cluster cannot be selected, so the Hyper-V Replica Broker fills in to represent all of the servers in the cluster.
• Hyper-V Replication from WSFC cluster1 to WSFC cluster2 was enabled from the VM in Failover Cluster Manager
  • Kerberos authentication (http) can be used for clusters in the same domain.
  • Certificate-based authentication (https) can be used for clusters in the same domain or in different subnets.

VM Migration

• VMs can be moved across clusters using the Planned Failover feature of Replication (this was confirmed in testing)
  • VMs must be stopped first since live migration is not possible.
• VMs can be live-migrated to another node within the same cluster from Failover Cluster Manager (also confirmed in testing)

Implementing Cluster-Set by EC (application)

1. ECX 2-cluster one shared disk

ECX 2 cluster note 2022.09.28

Reproduce the same configuration as Cluster-Set only with ECX

Testing Configuration

• Two clusters of 2 nodes each were created.
• They connect to the same iSCSI and check the contents of the disk.
• Set start script and end script to control VM.
• I stop the disk resource from one of the cluster, to prevent seeing on the same disc at the same time.
• I use RestfulAPI in my custom monitor resource script to check if other clusters are running
• It is difficult to FO to other cluster with only script resource.
  • There is a risk that the script will not run when the server turned off at the same time by using the script resource.
  • It is not possible to determine when to move manually and when not to move during maintenance etc.

VM Migration

• VMs can be moved across clusters using the script resource and Stopping the disk resource by manual.

ECX 2 cluster script outline

• This script can be started on any server
• The information of cluster1, cluster2, server1, server2, server3, server4 as variables
• specify as an array
• subsequent processing
• Get the status of the cluster to which the server belongs, change the URL later

$result = curl.exe -u root:cluster-0 http://127.0.0.1:29009/api/v1/groups/failover1 --noproxy 127.0.0.1 | ConvertFrom-Json
if ($result.groups.status -eq "Online")
{
    Write-Output "Failover group is running on my cluster."
    exit 0
}

• Get group status of other cluster
• Insert the process to get the status of the group from other clusters here
• If a communication error occurs, go to another server in another cluster to get the status

if ($result.groups.status -eq "Online")
{
    Write-Output "Failover group is running on the other cluster."
    exit 0

Start group with RESTful API or clprexec

2. ECX 2 cluster two shared disks with Hyper-V Replication

This Cluster Set type solution uses Hyper-V replication between active nodes on each cluster. If required, a cross-cluster VM “move” can be done using the Planned Failover command and then reversing replication.

Notes:

• All nodes in each cluster belong to the same domain.
• A cross-cluster VM move is not automated. It has to be done manually.
• Scripts automate intra-cluster failovers, changing Hyper-V Replica’s Primary or Replica servers as needed.
• Kerberos authentication was used.
• RESTful API used to query for the active server of the other cluster.

Testing Configuration

• Two ECX clusters of two nodes each were created (Windows Server 2019).
• Each cluster has one iSCSI shared disk.
• FIP resource is used.
• Script resource is used.

Link to more specific details.

Other

Move-VM

If you only want to temporarily move a VM to another cluster (while performing maintenance), and do not care about HA or DR, this PowerShell command will work. The command can Live-Migrate a VM from a node in Cluster-A to a node in Cluster-B.

Move-VM -Name <vm name> -DestinationHost <active node in other cluster> -IncludeStorage -DestinationStoragePath <location for files>