Storage
Virtual Machine Orchestrator (VMO) manages the storage that backs VM disks. From Infrastructure > Storage, you can manage StorageClasses, storage pools, and DataVolumes. PaletteAI VM Launchpad ships with Piraeus/LINSTOR as the default storage backend, but VMO works with any StorageClass that supports dynamic provisioning.
Storage Providers
VMO does not require a specific storage backend. It works with any Kubernetes StorageClass that supports dynamic provisioning.
The appliance backend depends on the appliance variant you install:
- Piraeus/LINSTOR provides replicated block storage for VM disks, StorageClass-based provisioning, and LVM-based storage pools. This backend ships with both the FIPS and non-FIPS Piraeus variants.
- Portworx provides enterprise distributed storage for VM disks. This backend ships with the non-FIPS Portworx variant.
You can also use other providers, such as host-path or Rook-Ceph, depending on the cluster configuration.
Set up the storage backend before you create a StorageClass. On the Piraeus/LINSTOR variants, VM storage is provisioned through LINSTOR storage pools. On the Portworx variant, create the Storage Cluster first. Create your StorageClasses on the configured backend afterward.
StorageClasses
StorageClasses define how VMO provisions PersistentVolumeClaims (PVCs). From Infrastructure > Storage, you can perform the following StorageClass operations.
| Operation | Description |
|---|---|
| List | View all StorageClasses in the cluster. |
| Create | Create a new StorageClass, when the underlying provider supports it. |
| Delete | Remove a StorageClass. If any PVCs or DataVolumes use the StorageClass, VMO blocks the deletion and lists the dependent volumes. |
| Set default | Mark one StorageClass as the cluster default. New PVCs that do not specify a StorageClass use the default. |
| Enable for VM Workloads | Make an existing StorageClass available for VM disks and volumes. The VM Workloads column shows which StorageClasses are enabled. |
Enable a StorageClass for VM Workloads
VMs can use only StorageClasses that are enabled for VM workloads. On the Portworx variant, Portworx creates its own CSI
StorageClasses, such as px-csi-db, when the Portworx storage cluster is running. To make a new
StorageClass available for VMs instead, create a StorageClass with Allow for VMs selected.
-
From the VMO left main menu, select Infrastructure > Storage > Storage Classes.
-
Check the VM Workloads column of the StorageClass. If the StorageClass is already enabled, you do not need to continue.
-
Open the detail page of the StorageClass, or right-click the StorageClass, and then select Enable for VM Workloads.
Create a StorageClass
-
From the VMO left main menu, select Infrastructure > Storage > Storage Classes.
-
Select Create Storage Class.
-
Configure the following fields.
Field Description Class Name The StorageClass name. Follow Kubernetes naming rules (lowercase, alphanumeric, hyphens); up to 128 characters. Default Class Select to mark this StorageClass as the cluster default. New PVCs that do not specify a StorageClass use the default. Allow Expansion Select to let PVCs expand. The underlying provider must support volume expansion. Allow for VMs Select to make this StorageClass available for VM workloads. Create StorageProfile for CSI-assisted cloning Select to enable offloaded DataVolume clones ( csi-clone) on Block volumes for this StorageClass.Reclaim Policy The Kubernetes reclaim policy: DeleteorRetain. Controls what happens to a PV when its PVC is released.Binding Mode ImmediateorWaitForFirstConsumer. Controls when volume binding and dynamic provisioning happen. -
Under Parameters, configure how VMO provisions volumes. The available parameters depend on the storage provider.
- Piraeus/LINSTOR
- Portworx
Field Description Select a Policy Select a Storage Policy to fill in the parameters below, or choose No Policy (manual parameters) to enter them manually. Storage Pool The name of the LINSTOR storage pool to provision from. Placement Count The number of replicas for each volume. Resource Group The LINSTOR resource group name. Advanced Parameters (Optional) Expand to configure other provider-specific parameters as key-value pairs. Portworx StorageClasses expose the Portworx volume options. The provisioner determines the full set of available parameters. The following options are the most common.
Field Description Replication Factor The number of synchronous volume replicas across nodes (Portworx repl), typically1,2, or3.IO Profile The Portworx IO profile ( io_profile) that tunes the volume for the workload, such asautoordb_remote.IO Priority The relative IO priority ( io_priority):high,medium, orlow.Filesystem The volume filesystem ( fs), such asext4orxfs.Encryption Enable Portworx volume encryption ( secure).Shared v4 (RWX) Enable Portworx shared volumes ( sharedv4). Keep it set tofalse, the default, forBlockvolume mode, which VM disks use. Portworx sharesReadWriteManyvolumes inBlockvolume mode without it, and they support live migration. Withtrue, Portworx rejects clones ofReadWriteOncevolumes in the class (sharedv4 does not match), so VMs that you create from images or templates do not finish provisioning. Portworx requirestrueonly forReadWriteManyvolumes inFilesystemvolume mode.Advanced Parameters (Optional) Expand to configure other Portworx parameters as key-value pairs. -
Select Create Storage Class.
Set the Default StorageClass
Only one StorageClass can be the cluster default at a time. New PVCs that do not specify a StorageClass use the default. Set the default in one of two ways:
- During creation: Select the Default Class checkbox in the Create Storage Class modal.
- On an existing StorageClass: Edit the StorageClass and select the Default Class checkbox.
StorageClasses are immutable in Kubernetes. Saving an edit deletes the existing StorageClass and recreates it with the updated settings. The same delete-and-recreate applies to the previous default when you mark a new one, because VMO clears the previous default in the same way. If any PVCs or DataVolumes reference the affected StorageClass, whether the one you are editing or the previous default when switching, the deletion step fails and the save is rejected. Delete or reassign the dependent volumes before editing or switching the default.
To set the default on an existing StorageClass, take the following steps.
-
From the VMO left main menu, select Infrastructure > Storage > Storage Classes.
-
In the Storage Classes list, select the row of the StorageClass you want to make the default. The details panel opens on the right.
-
Select Edit.
-
Select the Default Class checkbox.
-
Select Save.
Delete a StorageClass
-
From the VMO left main menu, select Infrastructure > Storage > Storage Classes.
-
In the Storage Classes list, select the row of the StorageClass you want to delete. The details panel opens on the right.
-
Select Delete Storage Class.
-
In the confirmation dialog, confirm the deletion.
If any PVCs or DataVolumes reference the StorageClass, VMO blocks the deletion and lists the dependent volumes. Delete or reassign the dependent resources before retrying.
Storage Profiles
Storage Profiles are CDI resources that define how VMO provisions DataVolumes for each StorageClass. They apply to any StorageClass, regardless of the storage backend. VMO auto-creates a StorageProfile when you enable Create StorageProfile for CSI-assisted cloning during StorageClass creation. The UI exposes only editing; VMO manages Storage Profile creation and deletion for you.
Edit a Storage Profile
-
From the VMO left main menu, select Infrastructure > Storage > Storage Profiles.
-
Select the StorageProfile you want to edit.
-
Configure the following fields.
Field Description Volume Mode BlockorFilesystem. For Piraeus/LINSTOR, we recommendBlockwithcsi-clone. ChooseFilesystemonly when the storage backend lacks Block volume support.Clone Strategy csi-clone,copy, orsnapshot. Controls how VMO clones DataVolumes backed by this StorageClass. -
Select Save.
Storage Pools
Storage pools apply to the Piraeus/LINSTOR variants. They are provider-specific constructs backed by native LINSTOR storage resources. VMO does not provide a dedicated Storage Pools tab; the underlying Piraeus/LINSTOR storage-pool APIs remain available and unchanged. To dedicate whole disks or partitions to a LINSTOR storage pool, use the disk partitioning workflow.
On the Portworx variant, you manage node-local devices through the Storage Cluster wizard instead.
Storage Clusters
Only the Portworx storage backend uses a Kubernetes StorageCluster object. The storage-cluster operations available to
you depend on the appliance variant you install.
- Piraeus/LINSTOR
- Portworx
Piraeus/LINSTOR appliances do not use a StorageCluster object and do not include a Portworx Storage Clusters tab.
Piraeus provisions VM storage through LINSTOR storage pools. To dedicate whole disks or partitions to a storage pool,
use the disk partitioning workflow and reference the resulting devices in your LINSTOR
configuration. Refer to Storage Pools for more information.
On appliances that use the Portworx storage backend, the Storage page includes an extra Portworx Storage
Clusters tab that does not appear on Piraeus/LINSTOR appliances. The Portworx pack installs the Portworx operator
without deploying a cluster, so you create the Portworx StorageCluster with the New Portworx Storage Cluster
wizard.
The Portworx operator supports one storage cluster for each Kubernetes cluster, so the New Portworx Storage Cluster button is unavailable while a storage cluster exists. The VM Launchpad cluster must have three control plane nodes, or one control plane node and three worker nodes. When the storage cluster is running, Portworx creates its CSI StorageClasses. Refer to Enable a StorageClass for VM Workloads to make one of them available for VM workloads.
The Storage backend field of the wizard offers the following options:
-
Local disks (software-defined): Portworx claims block devices on each node.
-
Pure FlashArray: Portworx provisions cloud drives on an Everpure (formerly Pure Storage) FlashArray. You need the management endpoint and an API token of each FlashArray, and the cluster nodes must reach the FlashArray over iSCSI or Fibre Channel.
Create a Portworx Storage Cluster
On VMware vSphere VMs, set the disks that Portworx uses to non-rotational before you create the storage cluster. Otherwise, the storage cluster does not initialize. Refer to Portworx Storage Cluster Does Not Initialize on VMware vSphere VMs for instructions.
-
From the VMO left main menu, select Infrastructure > Storage > Portworx Storage Clusters.
-
Select New Portworx Storage Cluster.
-
Enter a Name. The Namespace is preset to
portworxand cannot be changed, because the Portworx Storage Cluster must use theportworxnamespace. -
(Optional) Under Metadata, add labels and annotations. The appliance pre-populates the Portworx annotations the cluster requires, such as
portworx.io/misc-args,portworx.io/pvc-controller-port, andportworx.io/pvc-controller-secure-port. -
In Storage backend, select Local disks (software-defined) or Pure FlashArray, and confirm that the Run on control plane configuration preset is enabled. If you select Pure FlashArray, the wizard adds a Pure FlashArray credentials panel and replaces the Storage and Nodes sections with Cloud Storage and Pure Platform.
-
(Pure FlashArray only) In the Pure FlashArray credentials panel, provide the connection details of each FlashArray. Portworx reads them from the
px-pure-secretSecret in theportworxnamespace.-
If the Secret does not exist, enter the Management endpoint and API token of each FlashArray, and then select Create secret. To add another array, select Add FlashArray.
-
If the panel reports that the Secret is present, you can change the stored credentials with Edit.
-
-
(Pure FlashArray only) In the Cloud Storage section, under Device Specs, set the Size in GiB of each pool. To add a pool, select Add pool. Keep the Provider set to
pure. -
(Pure FlashArray with Fibre Channel only) Portworx connects to the FlashArray over iSCSI by default. To use Fibre Channel, select the Env section, and then add the
PURE_FLASHARRAY_SAN_TYPEenvironment variable with the valueFC. -
Configure the remaining wizard sections from the left panel. The following table describes the most common fields. The Summary and Validation panels on the right update as you go.
Section What you configure General Secrets Provider: k8sstores Portworx credentials as Kubernetes secrets.
Custom Image Registry: a registry prefix prepended to every Portworx image, required for airgapped clusters and left empty on connected clusters to pull fromdocker.io.
Image: the Portworx image.Storage (Local disks only) The node-local block Devices Portworx claims, the System Metadata Device, and optional Journal Device, KVDB Device, and Cache Devices. Select Force Use Disks to reuse disks that already carry data, or Use All to claim every available disk. Use the block-device picker's disk partitioning workflow to carve dedicated KVDB, journal, or metadata partitions. Nodes (Local disks only) Per-node storage overrides for heterogeneous hardware. Cloud Storage (Pure FlashArray only) The size in GiB of each Device Specs pool, the Kvdb Device Spec and System Metadata Device Spec sizes, the optional Journal Device Spec and Capacity Specs, and the cloud drive Provider ( pure).Pure Platform (Pure FlashArray only) Optional Fusion and Integration Operator settings. Kvdb The Portworx key-value store. Keep Internal selected for an internal KVDB, or clear it and provide Endpoints and an Auth Secret for an external KVDB. Select Enable TLS to secure KVDB traffic. Csi Enable the Portworx CSI driver and choose an internal or external CSI deployment. Network The Data Interface Portworx uses for storage traffic and the Mgmt Interface it uses for management traffic. Volumes Default settings applied to new Portworx volumes. Autopilot Enable Portworx Autopilot to automatically expand capacity as pools fill. Delete Strategy How Portworx cleans up when you delete the cluster: Uninstall,UninstallAndWipe, orUninstallAndDelete. Select Ignore Volumes to leave provisioned volumes in place.Env Environment variables that the storage cluster passes to Portworx, such as PURE_FLASHARRAY_SAN_TYPE.If you use a V1 Storage Cluster, you do not need to create a separate metadata device. Portworx uses one unmounted partition or a raw unmounted disk that you provide.
-
(Optional) Select Advanced to edit the
StorageClusteras raw YAML instead of using the form. -
Review the Summary and confirm that Validation reports no issues, and then select Save in the upper-right corner.
Portworx claims and formats the disks and devices you select in the Storage section. Confirm that each disk carries nothing you need before you save the cluster.
Review a Portworx Storage Cluster
-
From the VMO left main menu, select Infrastructure > Storage > Portworx Storage Clusters.
-
Select a Portworx
StorageClusterto review its configuration, status, and cluster metrics such as nodes online, cluster size, capacity used, and capacity total. The Phase isInitializingwhile Portworx starts,Runningwhen the storage cluster is ready, andDegradedif the installation failed or while you delete the storage cluster.
Rotate Pure FlashArray Credentials
-
From the VMO left main menu, select Infrastructure > Storage > Portworx Storage Clusters.
-
Select Edit next to your storage cluster.
-
In the Pure FlashArray credentials panel, select Edit.
-
Update the Management endpoint or API token of each FlashArray. The form never displays stored tokens. Leave an API token blank to keep the stored token for that endpoint.
-
Select Rotate credentials.
-
Select Save in the upper-right corner. Portworx uses the new credentials without a restart.
Partition a Disk for Storage
VMO can partition a discovered disk on a cluster node directly from a device field, for example to carve a small KVDB, journal, or metadata partition for Portworx out of a larger disk, or to split a single disk into a metadata partition and a data partition.
You partition a disk from the block-device picker, the control that opens on device fields that support partitioning, such as the KVDB Device, Journal Device, and Metadata Device fields in the Portworx Storage Cluster wizard. Open the picker for one of these fields and select Partition a disk to launch the partition tool.
Prerequisites
-
Disk partitioning is enabled on the node agent. The feature is opt-in and dormant by default. Enabling it adds a container to the node-agent DaemonSet that runs with elevated privileges so it can write partition tables to raw block devices. Both partitioning and the Wipe disk action require this feature. When it is not enabled, the tool reports that disk partitioning is unavailable.
-
Your account holds the
vmo:storage:partitionpermission. Among the built-in roles, only Platform Admin holds it. This permission is separate fromvmo:storage:writebecause partitioning destructively rewrites the partition table on a physical host disk. -
The node agent is running on the target nodes. The workflow offers only devices from live node-agent discovery.
Author Partitions
Start from a device field that supports partitioning in the Portworx Storage Cluster wizard. Refer to Create a Portworx Storage Cluster to open the wizard.
-
Open the block-device picker for the device field and select Partition a disk.
-
Select the target node or nodes and one discovered device. When the field is scoped to specific nodes, the node scope is pre-filled. For a cluster-level field, select the nodes yourself, and VMO partitions the same device name on each.
-
Author the partitions. Each partition row takes a size in GiB and an optional reference label. Select Add partition to split one disk into several labeled partitions in a single operation, for example a 64 GiB metadata partition plus a data partition.
-
(Optional) To size the last partition to the remaining free space on the disk, set it to Use remaining space instead of a fixed size. Only the last row can use remaining space.
-
Review the summary, then type the exact device name to confirm. The confirmation is deliberate, because the operation writes a partition table to a physical disk.
-
Select Submit. VMO validates the request against current discovery data and records it as a
PartitionIntentresource. The node agent writes the partitions on the target node and reports the created partition paths.
The partitions you create become selectable in device pickers after the node agent's next discovery scan. The field you partitioned from is filled with the new partition path right away.
VMO never partitions or wipes the disk that hosts the operating system. The safety guard detects only active mounts, holders, and in-use state. It does not detect inactive members, such as an unmounted LVM physical volume or a non-assembled software RAID member. Before you partition or wipe a disk, confirm it carries nothing you need.
Wipe a Disk
Wipe disk is a separate, destructive action that clears a disk's entire partition table so you can repartition it from scratch. Use it to reset a disk whose existing partition table blocks a fresh partitioning attempt.
-
In the partition tool, select the target node or nodes and one device.
-
Select Wipe disk.
-
Type the exact device name to confirm.
VMO does not wipe a disk that hosts the operating system, or a disk whose partitions are mounted, in use, or have active holders.
Storage Policies
Storage Policies are reusable parameter presets for StorageClass creation. When creating a StorageClass, select a policy from the Select a Policy drop-down to fill in provider-specific parameters instead of entering them manually. VMO ships with two built-in policies:
-
Piraeus DRBD Performance Tuning, for Piraeus/LINSTOR workloads with live-migration support.
-
Portworx VM Workload Defaults, for Portworx VM disks. It sets Replication Factor to
3, IO Priority tohigh, Shared v4 (RWX) tofalse, and Filesystem toext4.
Create a Storage Policy
-
From the VMO left main menu, select Infrastructure > Storage > Storage Policies.
-
Select Create Policy.
-
Configure the following fields.
Field Description Policy Name The policy name. Follow Kubernetes naming rules (lowercase, alphanumeric, hyphens); up to 128 characters. Display Name A human-readable name for the policy. Shown in the Storage Policies list and in the Storage Class Select a Policy drop-down. Description (Optional) A short description of the policy. Provider The storage provider the policy applies to (for example, piraeusorportworx). -
Use Add Group and Add Parameter to define the parameters the policy sets. VMO applies these parameters when you select the policy during StorageClass creation. For a Portworx policy for
Blockvolumes, keepsharedv4set tofalse. -
Select Save.
Edit a Storage Policy
-
From the VMO left main menu, select Infrastructure > Storage > Storage Policies.
-
In the Storage Policies list, select the row of the policy you want to edit. The details panel opens on the right.
-
Select Edit Policy.
-
Adjust the parameter values in the Edit Policy modal.
-
Select Save.
Built-in policies (labeled with a Built-in tag) restrict edits to parameter values only. The policy name, description, and parameter structure are managed by the system. Policies you create yourself allow full edits.
Delete a Storage Policy
-
From the VMO left main menu, select Infrastructure > Storage > Storage Policies.
-
In the Storage Policies list, select the row of the policy you want to delete. The details panel opens on the right.
-
Select Delete Policy.
-
In the confirmation dialog, confirm the deletion.
Built-in policies cannot be deleted. The Delete Policy button is only available on policies you create yourself.
DataVolumes
DataVolumes are Containerized Data Importer (CDI) resources that back VM disks. VMO supports the following DataVolume sources.
| Source | Description |
|---|---|
| Upload | Upload from a browser. Uses the CDI upload proxy. |
| URL | Import from an HTTP or HTTPS URL. |
| Blank | An empty disk of the specified size. |
| Clone | Clone from an existing PVC or DataVolume. |
| Registry | Import from a container registry. |
VMO lists DataVolumes on Infrastructure > Storage, where you can create and delete them.
The DataVolumes UI does not include a resize option. To change a DataVolume's size, delete and recreate it at the desired size.
Create a DataVolume
-
From the VMO left main menu, select Infrastructure > Storage.
-
Select Create DataVolume.
-
Configure the common fields.
Field Description Name The DataVolume name. Follow Kubernetes naming rules (lowercase, alphanumeric, hyphens); up to 128 characters. Namespace The namespace where VMO creates the DataVolume. Storage Class The StorageClass that backs the DataVolume. Defaults to the cluster default (for example, vmo-default-sc).Size The DataVolume size. Enter a number and select a unit: Gi,Mi, orTi.Access Mode ReadWriteOnce,ReadWriteMany, orReadOnlyMany. The StorageClass must support the mode you select.Volume Mode BlockorFilesystem. -
Select a Source and configure its type-specific fields. For Upload, URL, and Registry sources, use the Image checkbox to control how VMO treats the volume. Leave the checkbox cleared for installer media, which VMO attaches as a CD-ROM drive in the VM. Select the checkbox for a bootable disk image or template source, which VMO clones as a VM boot disk.
- Upload
- URL
- Blank
- Clone
- Registry
Upload a local file through the CDI upload proxy.
Field Description File Choose a local file. Accepts .iso,.img, and.qcow2files.Import from an HTTP or HTTPS URL.
Field Description Source URL The HTTP or HTTPS URL to the disk image. Create an empty disk of the size specified in the common fields. Useful for boot disks or scratch volumes.
Clone an existing PVC or DataVolume.
Field Description Source Namespace The namespace that contains the source. Source PVC The PVC to clone from. Import from a public container registry.
Field Description Image Reference Public container image reference, for example, docker://quay.io/org/image:tag. VMO adds thedocker://prefix automatically if you omit it.infoThe VMO UI does not include an option to add private registries that require authentication. Contact Spectro Cloud Support for manual configuration steps.
Delete a DataVolume
-
From the VMO left main menu, select Infrastructure > Storage > Data Volumes.
-
Locate the DataVolume and choose one of the following actions:
- Select the trash icon in the row's Actions column.
- Check the boxes on the rows you want to delete, then select Delete N.
-
In the confirmation dialog, type the DataVolume name and select Delete.
Deletion is irreversible. Ensure you have backups or snapshots if the data is important.
Persistent Volume Claims
The VMO UI does not include operations for managing
PersistentVolumeClaims (PVCs) directly. Provision VM
storage using DataVolumes, which provide the user-facing abstraction; VMO manages the underlying PVCs
for you. To inspect or modify PVCs directly, use kubectl or another Kubernetes tool.
StorageClass Auto-Detection
VMO does not require a static StorageClass configuration. On startup, VMO reads the spec.storageClassName of its own
data PVC, such as vmo-manager-data, and caches the value. VMO uses the detected StorageClass as the default for:
- New VM disks.
- Golden image DataVolumes.
- Virtio PVCs.
- Builder disks.
If VMO does not detect a StorageClass, such as when no bound PVC exists, VMO falls back to the cluster default. When no cluster default exists, you must specify a StorageClass explicitly when creating resources.
To control the default storage for VM disks, configure the VMO data PVC with the StorageClass you want to use. This PVC is the single source of truth for default storage.