Building New HVM Clusters¶

Prerequisites¶

Before creating an HVM cluster, identify the target layout and ensure the following requirements are met:

Cluster Creation¶

1. Navigate to Infrastructure > Clusters

2. Click + Add Cluster

3. Select HVM as the cluster type

4. Select the required layout: HVM 1.3 for HVM OS/Ubuntu 24.04 or HVM 2.0 for HVM OS 26.04

5. Complete the following fields:

   Field	Description
   SSH Hosts	Comma-separated IP addresses of all hosts
   SSH Port	SSH port (default: 22)
   Username	SSH username with root or sudo access
   Password/Key	Authentication credential
   Data Device	Data device path (HCI only)
   Management Network Name	Network name for management traffic
   Storage Network Name	Network name for storage traffic
   Compute Network Name	Network name for compute traffic
   Overlay Network Name	Legacy/layout 1.3 underlay/tunnel network name used by the HVM networking plugin. Verify tunnel-interface reachability and MTU; layout 2.0 uses Virtual Switch networking instead
   Compute VLANs	VLAN IDs for compute networks
   CPU Architecture/Model	Processor architecture selection
   Witness	Distributed Worker witness for a two-node GFS2 cluster or stretch cluster. Configure it after cluster creation. For two-Host GFS2 with no Site Groups, see Two-Node Clusters with a Witness. For stretch clusters with Site Groups, see Stretch Clusters & Witness Nodes.

6. Click Complete to begin automated provisioning

Cluster Permissions and Provisioning Impact¶

From the Clusters list, open More > Permissions for a cluster to scope where that cluster can be used:

• Groups control which Infrastructure Groups can use the cluster as a provisioning target. A user also needs role access to the Group and the relevant provisioning features.

• Service Plans limit the plans that can be selected when provisioning to the cluster. For example, selecting only a 1 CPU, 2 GB plan means other plans are not offered after this cluster is selected, subject to the selected Instance layout and image requirements.

• Default retains the product’s default permission behavior rather than creating an explicit Group or Service Plan allow-list. It does not mean that this cluster is the preferred provisioning target and does not override role, layout, image, or capacity filtering. Use an explicit selection when the cluster must be restricted.

Permission changes affect subsequent wizard choices. They do not resize, move, or otherwise alter existing VMs. After changing permissions, test with a non-administrator role: start an Instance deployment, select the intended Group and HVM target, and confirm that only the allowed clusters and Service Plans are shown. If an expected plan is absent, also check that the plan is active, uses the HVM/KVM provision type, and satisfies the selected layout and Virtual Image minimums.

Automated Provisioning Phases¶

Once cluster creation is initiated, HPE Morpheus Software executes the following automated phases:

Phase 1: Package Installation¶

Installs required packages on all hosts:

• KVM/QEMU

• libvirt

• Corosync

• DLM

• GFS2 utilities

• Ceph libraries

• swtpm

• jq

• nfs-common

Phase 2: Host Networking¶

Configures the networking model for the selected layout:

• Layout 1.3: Creates OVS bridges for management, compute, and overlay networks and configures the libvirt Management network

• Layout 2.0: Creates the default Virtual Switch and applies Linux bridge networking through hvmcli

See HVM Networks for layout 1.3 networking or Virtual Switches for layout 2.0 networking.

For layout 2.0, complete the scenario checklist in Preparing HVM Hosts before creating Virtual Switches. VLAN IDs are optional: use explicit VLAN IDs with an upstream trunk for a decoupled design without native VLANs, or leave a segment untagged only when the upstream port is intentionally configured for that traffic.

Phase 3: Host Preparation¶

Prepares each host for cluster operation:

• Firewall configuration

• libvirt configuration

• IOMMU enablement

• Hugepages configuration

• Workqueue tuning

• Directory structure creation

Phase 4: Cluster Services¶

Establishes cluster communication and coordination:

• SSH key generation and distribution via Cypher

• Corosync authkey generation and distribution via Cypher

• Corosync configuration with knet transport

• DLM configuration with fencing enabled

• Service ordering and startup (corosync → DLM)

Phase 5: iSCSI/Storage¶

Configures storage connectivity:

• iSCSI initiator name assignment

• Multipath configuration

• OOM protection for storage services

Phase 6: Storage Pools¶

Creates local storage pools:

• Cloud-init ISO pool

• Local storage pool

Phase 7: Quorum Activation¶

Activates the agent-based quorum system:

• Server sends quorumInfo to all agents

• Agents persist cluster member data to .quorum-nodes

• Agents begin peer-to-peer quorum pinging

HPE Clustered Datastore (Shared LUN) Setup¶

The HPE Clustered Datastore (GFS2) provides shared storage accessible by all cluster hosts simultaneously.

Adding iSCSI Targets¶

1. Navigate to Infrastructure > Clusters > [Cluster] > Storage > iSCSI

2. Click Add

3. Enter the Target IP and Port (default: 3260)

4. Save the configuration

Creating the Datastore¶

Create the HPE Clustered Datastore from Infrastructure ‣ Storage ‣ Data Stores or from the HVM Cluster’s Datastores tab. See Storage Lifecycle for the canonical prerequisites, fields, resulting cluster operations, verification, and storage lifecycle procedures.

Configuring Heartbeat Datastore¶

1. Edit the shared datastore properties

2. Enable the Heartbeat Target option

Verification¶

After cluster creation, verify the cluster is healthy:

Corosync membership:

corosync-quorumtool -l

Confirm all expected compute Hosts are listed. Corosync provides membership to DLM but does not decide Agent quorum for layouts 1.3 and 2.0. Do not use the Corosync Quorate value as the HVM health decision, especially for two-node and stretch topologies.

DLM status:

systemctl is-active dlm
dlm_tool status -v

Confirm all nodes show member=1.

Agent quorum:

curl -k https://localhost:7443/quorum

UI verification:

Navigate to Infrastructure > Clusters > [Cluster] > Summary > Quorum panel and confirm:

• Quorum status: ACHIEVED

• All nodes: ONLINE

• Coordinator: assigned

• Lockspaces: OK

For a two-node GFS2 cluster, also confirm the Distributed Worker witness is listed and reachable. See Two-Node Clusters with a Witness.

Cluster Sizing¶