Host Maintenance & Evacuation¶

Overview¶

Host maintenance mode allows administrators to take a host offline for servicing (hardware repair, OS patching, firmware updates) without triggering unnecessary VM failover. When a host enters maintenance mode, HPE Morpheus Software evacuates VMs to other cluster hosts via live migration before marking the host offline.

This procedure evacuates a Host. To evacuate VM disks from a datastore that is being serviced or retired, use datastore maintenance mode in Storage Lifecycle.

Entering Maintenance Mode¶

Procedure¶

1. Navigate to Infrastructure > Clusters > [Cluster] > Hosts

2. Click the actions menu for the target host

3. Select Enter Maintenance Mode

4. HPE Morpheus Software begins the automated evacuation process

Automated Evacuation Process¶

When maintenance mode is initiated, HPE Morpheus Software performs the following:

1. Capacity validation (dry run) — Before moving any VMs, HPE Morpheus Software verifies that remaining hosts have sufficient memory to accommodate all VMs. The system ensures no host exceeds 95% projected memory utilization after receiving migrated VMs. If capacity is insufficient, maintenance mode is refused with an error.

2. VM iteration — VMs are processed in descending order by memory allocation (largest VMs first).

3. Target host selection — For each VM, HPE Morpheus Software selects a target host with the most available memory that can accommodate the VM without exceeding 95% utilization.

4. Live migration — Running VMs on shared storage are live-migrated (no downtime). Each migration is retried up to 3 times on failure, with a 15-second pause between attempts.

5. Cold migration — Powered-off VMs are cold-moved (XML definition relocated).

6. Status update — Once all eligible VMs are evacuated, the host status changes from maintenancing to maintenance.

VMs Skipped During Evacuation¶

The following VMs are not moved during maintenance evacuation:

For permanent decommissioning, maintenance mode is only the evacuation step. Complete the supported cluster removal and post-removal health checks in Adding and Removing Hosts.

VMs With Local Storage¶

VMs that have volumes on local datastores cannot be live-migrated. These VMs must be powered off before the host enters maintenance mode. If a running VM has local storage, maintenance mode will report a failure for that VM and continue with the remaining VMs.

To handle VMs with local storage:

1. Power off the VM manually before initiating maintenance mode

2. Or move the VM’s storage to a shared datastore before initiating maintenance mode

Capacity Requirements¶

Maintenance mode requires that remaining hosts can absorb the workload:

• Each target host must have available memory greater than the VM’s allocated memory

• Projected utilization on the target host must remain below 95%

• If no host has sufficient capacity for a given VM, the operation fails

Exiting Maintenance Mode¶

Procedure¶

1. Navigate to Infrastructure > Clusters > [Cluster] > Hosts

2. Click the actions menu for the host in maintenance

3. Select Exit Maintenance Mode

What Happens¶

• The host status changes from maintenance to provisioned

• The host’s maintenance mode flag is cleared

• Stat collection and quorum participation resume immediately

• VMs are not automatically moved back to the host

Planned vs. Unplanned Host Absence¶

Proper Shutdown Sequence¶

To cleanly shut down a host without triggering failover:

1. Enter maintenance mode (evacuates VMs)

2. Wait for evacuation to complete (host status shows maintenance)

3. Power off or reboot the host for servicing

Re-joining After Maintenance¶

When a host boots back up after maintenance:

1. The HPE Morpheus Software agent starts and reads .quorum-nodes from /opt/morpheus-node

2. The agent resumes peer-to-peer quorum pinging on port 7443

3. Other cluster agents detect the returning host as reachable

4. Quorum status updates to include the returning host

5. The host is available to receive VMs (via Dynamic Placement, manual migration, or new provisioning)

Maintenance During Cluster Updates¶

When performing rolling cluster updates (see Upgrading Clusters), HPE Morpheus Software automatically enters and exits maintenance mode for each host as part of the update process. A Dynamic Placement lock is acquired during maintenance to prevent resource scheduling from interfering with the update cycle.