HPE VM Clusters

An HPE VM cluster is a hypervisor clustering technology utilizing KVM. Beginning with just a few basic Ubuntu boxes, Morpheus can create a cluster of hypervisor hosts complete with monitoring, failover, easy migration of workloads across the cluster, and zero-downtime maintenance access to hypervisor host nodes. All of this is backed by Morpheus Tenant capabilities, a highly-granular RBAC and policy engine, and Instance Type library with automation workflows.

Features

Host Features

• Automated HPE VM cluster provisioning

• CEPH storage configuration for multi-node clusters

• CEPH summary, a high-level dashboard of CEPH components and status

• DRS, automatic rebalancing of clusters based on resource consumption

• Compatibility validation of network and storage devices at time of cluster provisioning

• Hypervisor console

• Configuration and deployment of OVS networks (VLANs)

• Cluster and individual host monitoring

• Add hosts to existing clusters

• Console support for cluster hosts

• Add, edit and remove networks and data stores from clusters

• Gracefully take hosts out of service with maintenance mode

• Migration of workloads across hosts

• Configurable automatic failover of running workloads when a host is lost

• Ability to add and provision to fibre channel storage resources or iSCSI storage resources via GFS2 or OCFS2 filesystem

• Integration with Morpheus costing

• Governance through Morpheus RBAC, Tenancy, and Policies

VM Features

• Workload provisioning and monitoring (Linux or Windows workloads)

• Console support for running workloads

• Affinity placement, pin VMs to hosts

• Brownfield discovery of existing VMs

• Reconfigure VM sizing

• Disk migration across datastores

• UEFI support

• Migration of VMs across hosts

• Configure automatic failover for individual VMs in the event a host is lost

• Reconfigure running workloads to resize plan, add/remove disks, and add/remove network interfaces

• Backup and restore HPE VM workloads, with optional synthetic full backups

• Clone VMs

• Take snapshots and revert to snapshots

• Morpheus library and automation support

• Integration with Morpheus costing features

Base Cluster Details

An HPE VM cluster using the hyperconverged infrastructure (HCI) Layout consists of at least three hosts. Physical hosts are recommended to experience full performance of the HPE VM cluster solution. In smaller environments, it is possible to create an HPE VM cluster with three nested virtual machines, a single physical host (non-HCI only), or a single nested virtual machine (non-HCI only) though performance may be reduced. With just one host it won’t be possible to migrate workloads between hosts or take advantage of automatic failover. Currently, a host must be a pre-existing Ubuntu 22.04 box with environment and host system requirements contained in this section. Morpheus handles cluster configuration by providing the IP address(es) for your host(s) and a few other details. Details on adding the cluster to Morpheus are contained in the next section.

Hardware Requirements

• Operating System: Ubuntu 22.04

• CPU: One or more 64-bit x86 CPUs, 1.5 GHz minimum with Intel VT or AMD-V enabled

• Memory: 4 GB minimum. For non-converged Layouts, configure HPE VM hosts to use shared external storage, such as an NFS share or iSCSI target. Converged Layouts utilize Ceph for clustered storage and require a 4 GB minimum memory per Ceph disk

• Disk Space: For converged storage, a data disk of at least 500 GB is required for testing. More storage will be needed for production clusters. An operating system disk of 15 GB is also required. Clusters utilizing non-converged Layouts can configure external storage (NFS, etc.) while Morpheus will configure Ceph for multi-node clusters

• Network Connectivity: HPE VM hosts must be assigned static IP addresses. They also need DNS resolution of the Morpheus appliance and Internet access in order to download and install system packages for HPE VM dependencies, such as KVM, Open vSwitch (OVS), and more

Note

Ubuntu 22.04 uses netplan for networking and it is the responsibility of the customer to establish recommended networking configurations prior to provisioning an HPE VM cluster. To configure a static IP address, change into the directory holding the config files (cd /etc/netplan) and edit the existing configuration file (/etc/netplan/50-cloud-init.yaml or /etc/netplan/00-installer-config.yaml or /etc/netplan/01-netcfg.yaml). If desired, backup the existing configuration prior to editing it (cp /etc/netplan/<file-name>.yaml /etc/netplan/<file-name>.yaml.bak). For additional information on configuration file formatting, refer to netplan documentation. Once the configuration is updated, validate and apply it (netplan try). The try command will validate the configuration and apply it if it’s valid. If invalid, it will automatically be rolled back.

Note

Clustered storage needs as much network bandwidth as possible. Network interfaces of at least 10 Gbps with jumbo frames enabled are required for clustered storage and for situations when all traffic is running through the management interface (when no compute or storage interface is configured). It’s highly likely that performance will be unacceptable with any lower configurations.

|mvm| Network Communication Ports

Description	Source	Destination	Port	Protocol
Morpheus Agent communication with the Morpheus appliance	HPE VM Host	Morpheus appliance server	443	TCP
HPE VM host configuration and management	Morpheus appliance server	HPE VM Host	22	TCP
HPE VM interhost communication for clustered deployments	HPE VM Host	HPE VM Host	22	TCP
Morpheus server SSH access for deployed virtual machines	Morpheus appliance server	HPE VM-hosted virtual machines	22	TCP
Morpheus server WinRM (HTTP) access for deployed virtual machines	Morpheus appliance server	HPE VM-hosted virtual machines	5985	TCP
Morpheus server WinRM (HTTPS) access for deployed virtual machines	Morpheus appliance server	HPE VM-hosted virtual machines	5986	TCP
Ceph Storage	HPE VM Host	HPE VM Host	3300	TCP
Ceph Storage	HPE VM Host	HPE VM Host	6789	TCP
Ceph MDS/OSD	HPE VM Host	HPE VM Host	6800-7300	TCP

Example Cluster Deployment

In this example cluster, each host box consists of:

• 4 vCPU

• 16 GB memory

• 20 GB OS boot disk

• 250 GB data disk (deployed to /dev/sdb)

• 3 network interfaces for management, storage, and compute traffic (set to eth0, eth1, and eth2, respectively, in this example. Your environment may differ.)

Note

250 GB data disks used in this example are simply for demonstration purposes. A typical test cluster should consist of at least 500 GB storage and more will be required for production. Do not raid disks on physical servers. Currently, only one data disk may be used, which is given in the DATA DEVICE configuration during cluster setup. In the very near future, an update will be provided to allow multiple data disks to be used. These will be added to the total Ceph storage in one large volume. Until that update, only one data disk may be given in the configuration.

HPE VM clusters must also live in Morpheus-type Clouds (See Infrastructure > Clouds). A pre-existing Morpheus Cloud may be used or a new Cloud could be created to handle HPE VM management.

Provisioning the Cluster

As mentioned in the previous section, this example is starting with three provisioned Ubuntu 22.04 boxes. I also have a Morpheus-type Cloud to house the cluster. Begin the cluster creation process from the Clusters list page (Infrastructure > Clusters). Click + ADD CLUSTER and select “HPE VM”.

Morpheus gives the option to select a hyperconverged infrastructure (HCI) LAYOUT or non-HCI. In this example, the HCI Layout is used (requires a three-node minimum). Next, configure the names and IP addresses for the host boxes (SSH HOST). The SSH HOST name configuration is simply a display name in Morpheus, it does not need to be a hostname. By default, configuration space is given for three hosts which is what this example cluster will have. You must at least configure one and it’s possible to add more by clicking the (+) button. The SSH PORT is pre-configured for port 22, change this value if applicable in your environment. Next, set a pre-existing user on the host boxes (SSH USERNAME and SSH PASSWORD) and SSH KEY. Use a regular user with sudo access.

In the next part of the modal, you’ll configure the storage devices and network interfaces. When Ceph initializes, it needs to be pointed to an initial data device. Configure this in the DATA DEVICE field. At this time, only one device may be given but in the near future, an update will allow for multiple devices to be configured which would be added to the total Ceph storage as one large volume. Find your disk name, if needed, with the lsblk command. In my case, the target device is located at /dev/sdb.

Though not strictly required, it’s recommended to have separate network interfaces to handle cluster management, storage traffic, and compute. In this example case, eth0 is configured as the MANAGEMENT NET INTERFACE which handles communication between the cluster hosts. eth1 is configured as the STORAGE NET INTERFACE and eth2 is configured as the COMPUTE NET INTERFACE. The COMPUTE VLANS field can take a single value (ex. 1) or a range of values (ex. 22-25). This will create OVS port group(s) selectable as networks when provisioning workloads to the cluster. If needed, you can find your network interface names with the ip a command.

Finally, only one CPU TYPE is currently supported (x86_64) though this may change in the future. For CPU MODEL configuration, we surface the entire database of model configurations from libvirt. If unsure or if you don’t know of a specific reason to choose one or the other, select host-model which is the default option.

At this point we’ve kicked off the process for configuring the cluster nodes. Drill into the Cluster detail page and click on the History tab. Here we can monitor the progress of configuring the cluster. Morpheus will run scripts to install KVM, install Ceph, install OVS, and to prepare the cluster. In just a short time, the cluster provisioning should complete and the cluster will be ready to deploy workloads.

Provisioning a Workload

At this point, the cluster is ready for workloads to be provisioned to it. The system default Ubuntu Instance Type contains a compatible Layout for HPE VM deployment. Add an Instance from the Instances list page (Provisioning > Instances). After selecting the Instance Type, choose a Group that allows for selection of the Morpheus-type Cloud containing the HPE VM cluster.

After moving to the next tab, select a Plan based on resource needs. From the RESOURCE POOL field, select the desired HPE VM cluster. When configuring VOLUMES for the new workload, note that space can be claimed from the Ceph volume. Within NETWORKS, we can add the new workload to one of the VLANS set up as part of cluster creation. Finally, note that we can choose the HOST the workload should run on.

Review and complete the provisioning wizard. After a short time, the workload should be up and running. With a workload now running on the cluster, we can take a look at some of the monitoring, migration, failover, and other actions we can take for workloads running on HPE VM clusters.

Monitoring the Cluster

With the server provisioned and a workload running, take a look at the monitoring and actions capabilities on the cluster detail page (Infrastructure > Clusters, then click on the new HPE VM cluster). View cluster performance and resource usage (Summary and Monitoring tabs), drill into individual hosts (Hosts tab), see individual workloads (VMs tab), and more.

Moving Workloads Between Hosts

To manually move workloads between hosts, drill into the detail page for the VM (from the VMs tab of the cluster detail page). Click ACTIONS and select “Manage Placement”. Choose a different host and select from the following placement strategies:

• Auto: Manages VM placement based on load

• Failover: Moves VMs only when failover is necessary

• Pinned: Will not move this workload from the selected host

Within a short time, the workload is moved to the new host.

Adding hosts

The process of adding hosts to a pre-existing cluster is very similar to the process of provisioning the cluster initially. The requirements for the new worker node will be identical to the nodes initially added when the cluster was first provisioned. See the earlier sections in this guide for additional details on configuring the worker nodes.

To add the host, begin from the HPE VM Cluster detail page (selected from the list at Infrastructure > Clusters). From the Cluster detail page, click ACTIONS and select “Add Worker”. Configurations required are the same as those given when the cluster was first created. Refer to the section above on “Provisioning the Cluster” for a detailed description of each configuration.

Once Morpheus has completed its configuration scripts and joined the new worker node to the cluster, it will appear in a ready state within the Hosts tab of the Cluster detail page. When provisioning workloads to this Cluster in the future, the new node will be selectable as a target host for new Instances. It will also be an available target for managing placement of existing VMs running on the cluster.

Note

It’s useful to confirm all scripts related to creating the new host and joining the new host to the cluster completed successfully. To confirm, navigate to the detail page for the new host (Infrastructure > Clusters > Selected Cluster > Hosts Tab > Selected Host) and click on the History tab. Confirm all scripts, even those run on the pre-existing hosts, completed successfully as it’s possible the new host was added successfully (green status) but failed in joining the cluster. When such a situation occurs it may appear adding the new host was successful though it will not be possible to provision workloads onto it due to not joining the cluster successfully.

Maintenance Mode

HPE VM cluster hosts can be easily taken out of service for maintenance when needed. From the host detail page, click ACTIONS and then click “Enter Maintenance.” When entering maintenance mode, the host will be removed from the pool. Live VMs that can be migrated will be moved to new hosts. VMs that are powered off will also be moved when possible. When a live VM cannot be moved (such as if it’s “pinned” to the host), the host will not go into maintenance mode until that situation is cleared. You could manually move a VM to a new host or you could power it down if it’s non-essential. After taking that action, attempt to put the host into maintenance mode once again. Morpheus UI provides a helpful dialog which shows you which VMs live on the host are to be moved as the host goes into maintenance mode. When maintenance has finished, go back to the ACTIONS menu and select “Leave Maintenance.”

Failover

HPE VM supports automatic failover of running workloads in the event of the loss of a host. Administrators can control the failover behavior through the “Manage Placement” action on any running VM. From the VM detail page, click ACTIONS and select “Manage Placement”. Any VM with a placement strategy of “Auto” or “Failover” will be eligible for an automatic move in the event its host is lost. When the loss of a host does occur, the workload will be up and running from a different cluster host within just a short time if it’s configured to be moved during an automatic failover event. Any VMs pinned to a lost host will not be moved and will not be accessible if the host is lost. When the host is restored, those VMs will be in a stopped state and may be restarted if needed.

This three-node cluster has three VMs running on the first host:

Each of these VMs is configured for a different failover strategy. When the host is lost, we should expect to see the first two VMs moved to an available host (since they have the “Auto” and “Failover” placement strategies, respectively). We should not see the third VM moved.

After loss of the host these three VMs were running on, we can see the lost host still has one associated VM in a stopped state. The other two VMs are running on a second host which is still available.

When the lost host returns, the moved VMs will come back to their original host. The third VM is associated with this host as well and is in a stopped state until it is manually restarted.

Adding an NFS Datastore

Existing NFS shares can be used with Morpheus HPE VM clusters for virtual machine storage. These are added and viewed from the Storage tab of the HPE VM cluster detail page and, once added and active, become selectable as targets for virtual machine storage.

Note

Ensure NFS is properly configured to allow all of the HPE VM hosts to access the shared directory, including permissions to read and write. For backup purposes, it’s also helpful to give Morpheus access to NFS.

Start by navigating to the Storage tab of the HPE VM cluster detail page. Make sure the Data Stores subtab is also selected. Here you will see a list of existing datastores with some additional information, such as type, capacity, and status. Click ADD. Enter the NAME for the datastore in Morpheus and select the TYPE as NFS Pool. Note that the datastore name cannot be changed once it has been created. This will update the available fields to include the additional fields needed to integrate the NFS server. Enter the SOURCE HOST which is the hostname or the IP address of the NFS server. Finally, enter the SOURCE DIRECTORY which is the directory path of the NFS share. Click SAVE.

Once the modal is saved, it will take a few minutes to initialize the new datastore and show a successful online status in Morpheus. Once this initialization process is completed, the datastore can now be used as VM storage for cluster.

Image Prep (Windows)

This section will go through the steps to prepare a Windows image which can be successfully provisioned to HPE VM clusters. Additionally, this image can serve as a template from which additional images and Morpheus Library items can be built. In this example case, we’ll start from downloading a Windows Server 2019 ISO directly from the Microsoft download center and go all the way through to creating a new Instance Type in Morpheus that users can provision on-demand.

With the Windows ISO already downloaded, begin by uploading the ISO as a Virtual Image in Morpheus. Virtual Images are added in Library > Virtual Images. Click + ADD and then choose “ISO.” Before adding the file itself, set the following configurations on the Virtual Image:

• NAME: A name for the Virtual Image in Morpheus, such as “Windows Server 2019 ISO”

• OPERATING SYSTEM: “windows server 2019”

• MINIMUM MEMORY: Filters out Service Plans at provision time which do not meet the minimum value. For this image type, I’ve set 4 GB

In addition to the above, there are a number of checkbox configurations here (many of them are in the expandable “Advanced” section), some of which are checked by default. They should all be unchecked except for “VIRTIO DRIVERS LOADED?” within the “Advanced” expandable section.

With the configurations set, it’s time to upload the ISO to Morpheus. Keep in mind that if you do not specify a bucket in which the file should be uploaded, it will be uploaded to the appliance itself. If you choose to do this, be sure you have enough space to store the images you need. Within the UPLOAD VIRTUAL IMAGE modal is a large dropzone labeled “Drop Files Here.” You can drag and drop the ISO file here or you can click the button labeled “Add File” and browse for it. A progress bar will appear, wait until the file is completely uploaded before you save and dismiss the modal. After the file has completely uploaded, click SAVE CHANGES.

Next, we’ll provision a VM from the ISO using the built-in HPE VM Instance Type. Once running, we will configure the VM to any specific requirements and convert it to a template. Navigate to Provisioning > Instances and click + ADD. On the TYPE tab of the Instance provisioning wizard, we select the Instance Type to provision. In this case, select “HPE VM” and click NEXT.

On the GROUP tab, select the Group and Cloud containing the target HPE VM Cluster and provide a name for the new Instance. In my case, I have an automatic naming policy setting my Instance name, but depending on your appliance configuration you may need to enter a custom name. Click NEXT.

On the CONFIGURE tab, first select the IMAGE. Select the Windows server ISO that was uploaded in the previous step. Based on the minimum memory configuration that was set on the Virtual Image, Plans which are too small will be filtered out. Among compatible Plans, select one that meets your requirements. Next, set the RESOURCE POOL, which is the HPE VM cluster you’re targeting. Configure disks and disk sizes, as well as network details (this will vary based on HPE VM cluster configuration). Finally, select the HOST, which is the HPE VM host within the cluster that the new Instance should initially be provisioned onto.

As a final step, we need to also expand the “Advanced Options” section and make sure “ATTACH VIRTIO DRIVERS” is checked. This will attach an ISO containing the VirtIO drivers which we’ll use later. Click NEXT.

The final two tabs of the wizard, AUTOMATION and REVIEW, do not require any configuration changes though you may want to review the Instance settings on the final tab. When done, click COMPLETE.

Click on the newly provisioning Instance from the Instances list page. Since this image is being provisioned for the first time, the image must be uploaded to the HPE VM host. This can take a little bit of time but any future attempts to provision workloads from this image will skip this step. Wait for the Instance to fully complete and appear in a green “Ready” status.

Once the Instance has fully finished provisioning, launch a console session by clicking ACTIONS and then “Open Console.” This will open a new window with a console session into the VM.

After selecting the language, click “Next.” On the following screen, click “Install Now.” This will begin the Windows setup process on our new VM. You’ll next select the operating system type you wish to install. For this example, I’m installing 2019 standard with desktop experience. Click “Next.”

Accept the licensing terms and click “Next.”

On the next screen, choose a custom install.

The next screen asks where Windows should be installed and may be empty. Click “Load Driver” to locate the mounted disk image containing the VirtIO drivers. The search should return a number of VirtIO SCSI controller packages for various Windows flavors. Select the proper package for the Windows version being installed. Click “Next.”

After a moment, we’re back at the screen asking where Windows should be installed. We should see the disk(s) of size and type selected at the time the VM was provisioned. Select the proper disk and click “Next.” The Windows installation will now begin. Once Windows has fully installed, proceed to the next step.

Following installation, Windows will restart and prompt for an Administrator user password. Set the password and log in as Administrator. Currently, there are no network interfaces configured. We need to install the VirtIO drivers to get this machine onto the network. We have a disk image mounted with the driver installer so we need to navigate to that drive and launch the installer. Open Windows Explorer and locate the drive in the side bar. In my case, it’s the E: drive. Right-click on virtio-win-guest-tools and select “Install.”

Step through the installer. Simply click “Next” or “Install” through each step, there are no configuration changes needed. Once the installer has completed, click “Finish.” You can confirm we now have a network interface by opening a Command Prompt session and using the ipconfig command. One network adapter should be listed.

We can now eject the two virtual disks, drives D: and E: in my case. Then, launch Windows Security so we can disable firewalls. Turn off firewall for domain, private network, and public network.

Next, back in Command Prompt, run winrm quickconfig to configure winrm. Within Services, ensure that winrm (Windows Remote Management) is set to automatic on startup. Right-click on the Start button and select Run. Enter “sysprep” and click OK. In the Windows Explorer window that appears, right-click on sysprep and click “Run as Administrator”. Under “Shutdown Options”, choose Quit and click OK. If this is set to shutdown, Morpheus will simply restart the VM. Once this is completed, a new file Sysprep_succeeded.tag appears in Windows Explorer.

We’re now done configuring Windows and the console window can be closed. We’ll move on to creating a template from the VM we just configured. Begin by opening an SSH session into the Morpheus appliance server. Confirm jq is up to date on the appliance box (apt install jq). Then, go ahead and stop the running Windows VM. We can do this from the Instance detail page in Morpheus. Click ACTIONS and then “Stop Server.” Still on the Instance detail page, click ACTIONS and then “Import as Image.” This will perform a snapshot and create a new Virtual Image (Library > Virtual Images).

The Virtual Image is not usable until it’s in an active status and the UI indication may display an active status even before it’s fully ready. If it’s “SAVING” or “QUEUED,” it is still being prepared and saved. To determine the current status of the Virtual Image, check with a call to Morpheus API like the one below. When the return output lists a status of “Active,” the image is ready to be provisioned from.

curl -k --request GET --url https://xx.xx.xx.xx/api/virtual-images/<id>
--header 'accept: application/json' --header 'authorization: Bearer xxx-xxx-xxx-xxx-xxx' |
jq '.virtualImage.status'

Once saved, additional configurations are needed on the Virtual Image in Morpheus. Edit the new Virtual Image and check the following configurations:

• MINIMUM MEMORY: Set as appropriate

• SYSPREPPED/GENERALIZED IMAGE?: Checked

• INSTALL AGENT?: Checked

• USERNAME: Remove if present

• PASSWORD: Remove if present

• VIRTIO DRIVERS LOADED?: Checked

All other checkbox-type configurations not mentioned in the above list should be unchecked. Click SAVE CHANGES.

At this point all image preparation steps are completed. Morpheus library items can now be created from this image by adding new Node Types, Layouts, and Instance Types. The complete steps for building a library item go beyond the scope of this particular guide but more detail on that process is available elsewhere in Morpheus UI documentation. Once the library items are created, new Instances may be provisioned complete with Morpheus Agent installed.

Hardware Passthrough

Morpheus includes hardware passthrough and pooling support for VMs provisioned to HPE VM Clusters. This support extends to GPU, USB and NVME devices. In order to be usable, such devices must be attached to a HPE VM Host or to multiple HPE VM Hosts. Morpheus includes controls to detach hardware devices from the underlying OS on the hypervisor host and add them to a pool where they are available for consumption by VMs. GPU hardware can be attached to VMs at provision time through custom Service Plans or on an ad-hoc basis by manually attaching hardware to existing VMs. Other supported hardware types must be past through to the VM after provisioning.

Creating Service Plans

When utilizing GPU hardware, it may be most convenient to use a custom Service Plan. This allows administrators to pre-configure the number of discrete hardware cards to attach to the VM at provision time. Once provisioned, the specified number of GPUs will be attached to the VM. For workloads which are provisioned and torn down regularly, this will save additional steps of manually attaching GPUs available from the pool. On teardown, attached devices of all types are released back to the pool. Though in some cases utilizing Service Plans may be most convenient, as you’ll see in the next sections it is not strictly required. It is also possible to manually attach GPUs or other hardware types on any existing VM runnning on HPE VM Clusters.

To create a new Service Plan, navigate to Administration > Plans & Pricing. Click + Add and then Service Plan. In the New Service Plan modal, first set the PROVISION TYPE configuration to “KVM.” This will update the available configuration fields and reveal the GPU COUNT configuration. Use this configuration to set the number of GPU hardware cards which should be attached to provisioned VMs using this Service Plan. All other configuration fields for Service Plans go beyond the scope of this guide. Consult Service Plans documentation for additional details specific to Service Plans.

Viewing and Assigning Hardware Devices

Hardware Devices are surfaced at the Host detail level and the VM (server) detail level through a Devices tab. The Host detail will show all attached hardware devices and their current status: Attached (to the host), detached (from the host), or assigned (to a VM). The VM detail will show only hardware devices currently attached to that VM. To begin consuming hardware devices with HPE VM Cluster VMs, look again at the list in the Devices tab on the Host detail page. Detach a piece of hardware from the host using the ACTIONS dropdown for that piece of hardware. As shown in the screenshot below, DETACH DEVICE is selected for a USB device.

Once detached, the status of the device is changed to “Detached” (see screenshot below) and the device is available for consumption by VMs running on this host.

To assign the hardware to a specific VM, click the ACTIONS dropdown once again for the now-detached hardware. Now, click ASSIGN DEVICE.

Within the ASSIGN DEVICE modal that will appear, select the server for device assignment and click EXECUTE.

The icon and status for the device in the hardware list has now changed to “Assigned.” If we then open a console session with this VM, we can see the USB device is assigned successfully and is usable by the guest OS.

The same process can be used to detach and assign GPU or NVME devices.