Upgrading Clusters¶

This section is the detailed HVM cluster upgrade reference (layout transitions, rolling Host updates, Agent behavior on the Host, and verification). For the overall HPE Morpheus Software upgrade entry point—including Manager VM, HA app nodes, and when to return here for HVM—start at Upgrading.

Upgrade sequence¶

Complete these steps in order. Layout and rolling cluster updates depend on a Manager that already includes the target release’s update definitions.

1. Upgrade the HPE Morpheus Software appliance (Manager) — Required before you can offer or run layout updates. See Step 1: Morpheus Appliance Upgrade below.

2. Run the cluster layout or rolling update — After the Manager is on the target version, perform the layout transition (for example 1.2 → 1.3) or a rolling host update from the cluster actions in the UI. See the sections that follow.

3. Confirm agent and quorum health — Host Agent upgrades that a layout update requires are handled automatically during the cluster update when needed; verify connectivity and Quorum afterward.

Step 1: Morpheus Appliance Upgrade¶

Upgrade the HPE Morpheus Software appliance before starting any HVM layout or rolling cluster update. The Manager release publishes the cluster update definitions and minimum Agent versions that the later steps consume. Do not attempt a layout update against an older Manager that does not yet offer that update.

How to upgrade the appliance¶

Follow the Getting Started upgrade hub for your topology — that is the canonical package procedure:

Upgrading

Primary path — HPE Morpheus Manager VM (Debian package): Most HVM deployments use the HPE Morpheus Manager QCOW2 (or equivalent) on Ubuntu/Debian. Use the Debian / Ubuntu section of Single Node/AIO Appliance Upgrades:

sudo morpheus-ctl stop morpheus-ui
sudo dpkg -i morpheus-appliance_x.x.x-1_amd64.deb
sudo morpheus-ctl reconfigure

Other topologies:

• Single-node on your own guest OS (.deb or .rpm): Single Node/AIO Appliance Upgrades

• 3-Node HA: 3-Node HA Upgrade

• Full HA: Full HA Upgrade

A Getting Started summary of Manager-then-HVM order is also in HVM Clusters After Manager Upgrade.

HVM behavior during the appliance upgrade¶

While the Manager is offline for the package install and reconfigure:

• Active HVM clusters continue to operate (agents are self-sustaining)

• Quorum decisions are made by agents independently of the appliance

• Heartbeat writes continue regardless of appliance connectivity

• VM failover still works during appliance downtime

Post-upgrade verification (before layout updates)¶

After the appliance upgrade completes and the UI is available:

1. Navigate to Infrastructure ‣ Clusters

2. Verify all clusters show their expected state

3. Check the Quorum panel for each cluster — confirm ACHIEVED status

4. Verify agent connectivity (all hosts show recent lastAgentUpdate)

Only after this verification, proceed to a layout upgrade or rolling cluster update.

Layout 1.2 → 1.3 Upgrade¶

The HVM 1.3 layout replaces the Pacemaker-based high availability stack with the HPE Morpheus Software agent’s QuorumCheckService. Clusters running layout 1.2 can be upgraded to 1.3.

What Changes¶

What Is Preserved¶

• Corosync — Continues to provide cluster membership and messaging (shared between both layouts)

• DLM — Continues to coordinate filesystem locks (shared between both layouts)

• HPE Clustered Datastores — GFS2 filesystems remain mounted and operational

• VMs — All running VMs continue without interruption

• iSCSI connectivity — Storage paths are unaffected

Upgrade Process¶

The layout upgrade is performed as part of a cluster update operation after the Manager is on a release that publishes the 1.2-to-1.3 update. During the upgrade:

1. Pacemaker services are disabled and removed

2. The HPE Morpheus Software agent’s QuorumCheckService activates

3. Quorum info is sent to all agents

4. Agents begin peer-to-peer quorum pinging

Post-Upgrade: Corosync Authkey¶

Rolling Cluster Updates¶

Layouts 1.3 and 2.0 support rolling updates where each host is updated one at a time without cluster downtime.

Update Process (Per Host)¶

For each host in the cluster, the rolling update performs:

1. Agent upgrade (if required) — If the host’s agent version is below the minimum required version, the agent is upgraded first. HPE Morpheus Software waits for the agent to reconnect after the upgrade.

2. Enter maintenance mode — VMs are evacuated from the host via live migration (see Host Maintenance & Evacuation). A Dynamic Placement lock is acquired to prevent resource scheduling from interfering.

3. Execute update scripts — Layout-defined scripts run on the host (package updates, configuration changes, etc.)

4. Exit maintenance mode — The host returns to normal operation and can receive VMs again. The Dynamic Placement lock is released.

5. Proceed to next host — The process repeats for the next host.

After all hosts complete the update phase:

6. Post-update scripts — Any scripts designated as postUpdate phase run on each host.

Failure Handling¶

Rollback Support¶

If the cluster layout defines rollback scripts, they are executed in reverse order on:

• The host where the failure occurred

• All hosts that previously completed the update phase successfully

This restores the cluster to its pre-update state.

Agent Package Updates¶

The HPE Morpheus Software agent on each cluster host can be updated independently of cluster layout updates. Automatic upgrades run when a layout or rolling cluster update requires a minimum Agent version. Manual upgrades are the supported path after Manager patch (and many minor) releases that do not ship an HVM layout update, so Hosts receive Agent improvements such as quorum and health-check logic and telemetry.

Automatic Agent Upgrades¶

During rolling cluster updates, if a host’s agent version is below the minAgentVersion required by the update definition, the agent is upgraded automatically before update scripts execute.

The automatic agent upgrade:

1. Triggers the agent upgrade on the host

2. Waits for the agent to disconnect and reconnect

3. Verifies the new agent version meets requirements

4. Proceeds with the update only after successful agent reconnection

Manual Agent Upgrades¶

Manual Host Agent upgrades are recommended when a Manager patch or minor release does not include an HVM layout or rolling cluster update. Layout-driven updates upgrade the Agent automatically only when minAgentVersion requires it. Patch releases often still ship Host Agent improvements—for example quorum and health-check logic, fencing behavior, and telemetry—that do not ride a layout change. After those Manager upgrades, trigger Agent upgrades on each HVM Host so the cluster picks up the new Agent package.

To upgrade from the UI:

1. Open the Host detail page (Infrastructure ‣ Clusters ‣ [Cluster] ‣ Hosts, or the Hosts list)

2. Expand ACTIONS and select Upgrade Agent

3. Wait for the Agent to disconnect and reconnect, then confirm the reported Agent version

Repeat for every HVM Host in the cluster. Alternatively, select Download Agent Script, connect to that Host over SSH, and run the downloaded script. Scripts are Host-specific; download and run the correct script on each Host.

After a Manager upgrade that does not include a layout update, check for available Host Agent upgrades before treating the environment as fully current. See also the Agent guidance in Release Lifecycle and the release notes for the Manager version you installed.

Pre-Upgrade Checklist¶

Before performing a cluster layout or rolling update:

Post-Upgrade Verification¶

After completing a cluster upgrade:

1. Quorum health:

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

   Confirm quorum is ACHIEVED with all nodes ONLINE.

2. Corosync membership:

   corosync-quorumtool -s
   

   Confirm Quorate: Yes and all expected nodes are listed.

3. DLM status:

   dlm_tool status -v
   

   Confirm all nodes show member=1.

4. Datastore mounts:

   mount | grep gfs2
   

   Confirm all expected HPE Clustered Datastores are mounted.

5. UI verification:

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

   • Quorum status: ACHIEVED

   • All nodes: ONLINE

   • Coordinator: assigned

   • Lockspaces: OK

6. VM health:

   Verify all VMs are running and accessible after the upgrade completes.