This plugin enables the integration of Pure Storage arrays with Proxmox Virtual Environment (VE) using multipath iSCSI or Fibre Channel (FC). It allows you to use Pure Storage as a backend for your virtual machine disks, and optionally for LXC container root volumes when configured, providing high performance and reliability.
- Features
- Linux containers (LXC)
- Prerequisites
- Installation
- Configuration
- Troubleshooting
- Contributing
- Easily enable and configure multipathing iSCSI to the Pure Array
- Active Cluster support (Experimental) - Automatic volume connection
on both arrays in Active Cluster configuration
- Volumes are automatically connected to hosts on both primary and secondary arrays
- Ensures high availability and optimal connectivity in Active Cluster setups
- Storage based snapshots
- Snapshots are presented in Proxmox like any other native Snapshot to a VM
- Snapshots are created by the Pure Array, making them deduped and instant
- Protection-group snapshot sync (QEMU)
- Scheduled or manual Pure protection-group snapshots (e.g.
pgroup-auto.277) are imported into the VM snapshot tree as sanitized names (e.g.pgroup-auto-277) - The snapshot description keeps the exact Pure parent and per-disk member names for rollback
- Opt-in via
pgroup_sync(default off). When enabled, sync is forked from storagestatus()on a configurable interval (pgroup_sync_interval, default 60s, minimum 30s) so pvestatd is not blocked - Imported entries are read-only for delete in Proxmox — remove them on the
array. LXC/
rootdirvolumes are detected and skipped outright (only QEMU VM configs are synced)
- Scheduled or manual Pure protection-group snapshots (e.g.
- Instant storage migration
- The plugin will automatically map the iSCSI volumes needed on the host the VM is being migrated to
- Optional LXC support — include
rootdirincontentto allow raw image-backed container root disks on the same block path as QEMU disks (see Linux containers (LXC))
The plugin exposes the Proxmox rootdir content type so a storage
section can host LXC root filesystems as raw volumes, alongside
images (QEMU disks), when you set e.g. content images,rootdir.
LXC on this backend uses the same allocation and device path logic as VM
disks. Support is best-effort and community-tested — validate create,
resize, and delete for your workloads on a real array before relying on it
in production. If you only need QEMU VMs, keep content images (omit
rootdir).
Before installing and using this plugin, ensure that your Proxmox VE environment meets the following prerequisites.
To ensure correct operation with Pure Storage, you need to configure your multipath settings appropriately. Specifically, you need to set find_multipaths to no in your multipath.conf file. This setting disables the automatic detection of multipath devices, which is necessary for Pure Storage devices to be correctly recognized.
Below is an example of how your multipath.conf file should look when configured for Pure Storage arrays:
defaults {
polling_interval 2
find_multipaths no
}
devices {
device {
vendor "PURE"
product "FlashArray"
path_selector "queue-length 0"
hardware_handler "1 alua"
path_grouping_policy group_by_prio
prio alua
failback immediate
path_checker tur
fast_io_fail_tmo 10
user_friendly_names no
no_path_retry 0
features "0"
dev_loss_tmo 60
recheck_wwid yes
}
}
blacklist {
device {
vendor ".*"
product ".*"
}
}
blacklist_exceptions {
wwid "3624a9370.*"
device {
vendor "PURE"
}
}
Initiate iSCSI according to the Proxmox Guidelines.
sudo iscsiadm -m discovery -t sendtargets -p <PURE ISCSI ADAPTER IP>
sudo iscsiadm -m node --op update -n node.startup -v automaticCaution
As long as there are no hostX entries in /sys/class/iscsi_host/ the plugin is not ready to be used.
There are two methods to install the plugin: manual installation and Debian package installation via APT.
Important
If you are using a cluster setup, install the plugin on all nodes in the cluster. The storage configuration will be automatically synchronized via corosync, but the plugin code must be present on each node.
Manual installation is useful for development or when you want to install from source.
sudo apt-get update
sudo apt-get install -y \
libwww-perl \
libjson-perl \
libjson-xs-perlgit clone https://github.com/kolesa-team/pve-purestorage-plugin.git
cd pve-purestorage-plugin# Create the custom plugin directory
sudo mkdir -p /usr/share/perl5/PVE/Storage/Custom
# Copy plugin file
sudo cp PureStoragePlugin.pm /usr/share/perl5/PVE/Storage/Custom/PureStoragePlugin.pm
# Set correct permissions
sudo chmod 644 /usr/share/perl5/PVE/Storage/Custom/PureStoragePlugin.pmsudo systemctl restart pvedaemon.service pveproxy.servicepvesm status
# The purestorage type should now be availableInstalling via Debian package is the recommended method as it handles dependencies automatically and provides easy updates.
Replace <PACKAGE_VERSION> with the desired version (e.g., 0.0.1). Check
the releases page
for available versions.
PACKAGE_VERSION="<PACKAGE_VERSION>"
wget "https://github.com/kolesa-team/pve-purestorage-plugin/releases/download/v${PACKAGE_VERSION}/libpve-storage-purestorage-perl_${PACKAGE_VERSION}-1_all.deb"sudo apt install ./libpve-storage-purestorage-perl_${PACKAGE_VERSION}-1_all.debdpkg -l | grep libpve-storage-purestorage-perl
# Should show the installed package version# Download new version
PACKAGE_VERSION="<NEW_VERSION>"
wget "https://github.com/kolesa-team/pve-purestorage-plugin/releases/download/v${PACKAGE_VERSION}/libpve-storage-purestorage-perl_${PACKAGE_VERSION}-1_all.deb"
# Upgrade
sudo apt install ./libpve-storage-purestorage-perl_${PACKAGE_VERSION}-1_all.debsudo apt remove libpve-storage-purestorage-perlTip
If you are using a cluster setup - this step needs to be executed only on one node of the cluster - corosync will do the rest for you.
After installing the plugin, you need to configure Proxmox VE to use it.
Since Proxmox VE does not currently support adding custom storage plugins
via the GUI, you will need to open shell and use pvesm command to add it:
pvesm add purestorage <storage_id> \
--nodes <proxmox_node_list> \
--address \
https://<purestorage_fqdn_or_ip> \
--token <purestorage_api_token> \
--vgname <purestorage_volume_group_name> \
--hgsuffix <purestorage_host_suffix>
--content imagesTo allow LXC container root disks on the same pool as well, use
--content images,rootdir instead of --content images (see
Linux containers (LXC)).
Alternatively, you can manually edit the storage configuration file
/etc/pve/storage.cfg.
purestorage: <storage_id>
nodes <proxmox_node_list>
address https://<purestorage_fqdn_or_ip>
token <purestorage_api_token>
vgname <purestorage_volume_group_name>
hgsuffix <purestorage_host_suffix>
content images
| Parameter | Description |
|---|---|
| storage_id | The storage identifier (name under which it will appear in the Storage list) |
| nodes | (optional) A comma-separated list of Proxmox node names. Use this parameter to limit the plugin to specific nodes in your cluster. If omitted, the storage is available to all nodes. |
| address | The URL or IP address of the Pure Storage API endpoint. Ensure that the Proxmox VE nodes can reach this address over the network. For high availability or Active Cluster configuration (experimental), you can specify multiple arrays separated by commas (e.g., https://array1.example.com,https://array2.example.com). When multiple arrays are specified, the plugin automatically connects volumes to hosts on all arrays. |
| token | The API token used for authentication with the Pure Storage array. This token must have sufficient permissions to create and manage volumes. For multiple arrays, specify tokens separated by commas in the same order as addresses. Each token must have permissions for its corresponding array. |
| vgname | (optional, conflicts with podname) The volume group name where virtual disks will be stored. This should match the configuration on your Pure Storage array. |
| podname | (optional, conflicts with vgname) The pod name where virtual disks will be stored. This should match the configuration on your Pure Storage array. |
| vnprefix | (optional) The prefix to prepend to name of virtual disks. |
| hgsuffix | (optional) A suffix that is appended to the hostname when the plugin interacts with the Pure Storage array. This can help differentiate hosts if necessary. |
| content | Specifies the types of content that can be stored. Use images for QEMU VM disks. Add rootdir (e.g. images,rootdir) to allow LXC root volumes on the same pool; see Linux containers (LXC). |
| protocol | (optional, default is iscsi) Specifies the storage protocol (iscsi, fc). |
| check_ssl | (optional, default is no) Verify the server's TLS certificate. Set to yes to enable SSL certificate verification. |
| token_ttl | (optional, default is 3600) Session token time-to-live in seconds. The plugin caches PureStorage API session tokens in /etc/pve/priv/purestorage/ (automatically replicated across cluster nodes). Tokens are proactively refreshed at 80% of TTL to prevent expiration during operations. |
| debug | (optional, default is 0) Enable debug logging. Levels: 0=off, 1=basic (token operations, main calls), 2=verbose (HTTP details, validation), 3=trace (all internals). Environment variable PURESTORAGE_DEBUG can be used as fallback when debug is not set in config. |
| pgroup_sync | (optional, default is no) Import Pure protection-group snapshots into QEMU VM snapshot trees. Off by default. See Protection-group snapshot sync. |
| pgroup_sync_interval | (optional, default is 60) Minimum seconds between protection-group snapshot sync runs (minimum 30). Only used when pgroup_sync is enabled. |
NOTE: Ensure that the token and other sensitive information are kept secure and not exposed publicly.
Example Configuration:
Single Array:
purestorage: pure
address https://purestorage.example.com
token abc123
vgname pure_vg
hgsuffix ""
content images
Active Cluster (Multiple Arrays) - Experimental:
purestorage: pure-cluster
address https://array1.example.com,https://array2.example.com
token token1,token2
vgname pure_vg
content images
LXC-capable pool (images and rootdir; best-effort — see
Linux containers (LXC)):
purestorage: pure-lxc
address https://purestorage.example.com
token abc123
vgname pure_vg
hgsuffix ""
content images,rootdir
Note
When multiple arrays are specified (Active Cluster configuration - experimental feature), the plugin automatically connects volumes to hosts on both arrays. This ensures high availability - if one array fails, volumes remain accessible through the other array. The plugin handles connection management on all arrays transparently.
Pure protection-group snapshots (for example pgroup-auto.277) are not
imported into the Proxmox VM snapshot tree unless you turn this on. The
feature is QEMU-only; LXC/rootdir volumes are skipped.
Enable it on a storage:
pvesm set <storage_id> --pgroup_sync 1
Optional interval (seconds, minimum 30, default 60):
pvesm set <storage_id> --pgroup_sync_interval 60
Or in /etc/pve/storage.cfg:
purestorage: pure
address https://purestorage.example.com
token abc123
vgname pure_vg
content images
pgroup_sync 1
pgroup_sync_interval 60
Imported names replace . with - (for example pgroup-auto-277). Delete
those snapshots on the Pure protection group, not from the Proxmox UI.
If you encounter issues while using the plugin, consider the following steps:
The plugin provides detailed debug logging to help diagnose issues. Debug output is written to syslog and can be viewed in Proxmox logs.
Enable debug logging:
Persistent (via configuration):
pvesm set <storage_id> --debug 1Temporary (for single command, when debug is not set in config):
PURESTORAGE_DEBUG=1 pvesm list <storage_id>Note: If
debugis set in storage configuration, it takes priority overPURESTORAGE_DEBUGenvironment variable.
Debug levels:
0- Off (production, default)1- Basic (token operations, main function calls, volume operations)2- Verbose (HTTP requests, token validation, API responses)3- Trace (all internal operations, detailed flow)
Example debug output:
PURESTORAGE_DEBUG=1 pvesm list pure-n1
Debug :: activate_storage (pure-n1)
Debug :: list_images (pure-n1, vmid=all)
Debug :: Read token cache from: /etc/pve/priv/purestorage/pure-n1_array0.json
Debug :: Token is valid (age: 125s)
Debug :: Using cached token from file (age: 125s)Common debug scenarios:
Debug volume creation:
PURESTORAGE_DEBUG=2 pvesm alloc <storage_id> <vmid> <volname> 10GDebug volume deletion:
PURESTORAGE_DEBUG=2 pvesm free <storage_id>:<volname>Debug API authentication issues:
PURESTORAGE_DEBUG=3 pvesm status <storage_id>View debug logs:
Check Proxmox daemon logs:
journalctl -u pvedaemon -fFilter for PureStorage plugin messages:
journalctl -u pvedaemon | grep -E "(Debug ::|Info ::|Warning ::|Error ::)"Check token cache status:
ls -lah /etc/pve/priv/purestorage/
cat /etc/pve/priv/purestorage/<storage_id>_array0.json | jq .Ensure that the Proxmox VE services are running correctly. You can restart the services if necessary:
sudo systemctl restart pve-cluster.service pvedaemon.service pvestatd.service pveproxy.service pvescheduler.serviceMultipath diagnostics:
List all multipath devices with details:
multipath -llVerbose multipath debugging:
multipath -ll -v3Show only Pure Storage devices:
multipath -ll | grep -A 10 "3624a9370"Reload multipath configuration:
systemctl reload multipathdiSCSI diagnostics:
List all iSCSI sessions:
iscsiadm -m sessionShow detailed session information:
iscsiadm -m session -P 3List all discovered targets:
iscsiadm -m nodeRescan iSCSI sessions:
iscsiadm -m session --rescanPureStorage volume diagnostics:
List all Pure Storage mapped devices:
ls -l /dev/mapper/3624a9370*Show device information:
lsblk | grep "3624a9370"Check device WWIDs:
/lib/udev/scsi_id --whitelisted --device=/dev/mapper/3624a9370<wwid>Show device mapper table:
dmsetup tableShow device mapper dependencies:
dmsetup deps -o devnameList all device mapper devices:
dmsetup ls --treeShow detailed info for specific device:
dmsetup info /dev/mapper/3624a9370<wwid>Partition management (kpartx):
List partitions on a device:
kpartx -l /dev/mapper/3624a9370<wwid>Add partition mappings:
kpartx -a /dev/mapper/3624a9370<wwid>Remove partition mappings:
kpartx -d /dev/mapper/3624a9370<wwid>Sync partition table:
kpartx -u /dev/mapper/3624a9370<wwid>Storage plugin diagnostics:
List all volumes on storage:
pvesm list <storage_id>Show storage status:
pvesm status <storage_id>Scan for new volumes:
pvesm scan <storage_id>Test volume allocation (dry-run):
pvesm alloc <storage_id> <vmid> test-volume 1G
pvesm free <storage_id>:test-volumeNetwork connectivity:
Test API endpoint connectivity:
curl -k https://<array_address>/api/2.30/arraysTest with API token:
curl -k -X POST https://<array_address>/api/2.30/login \
-H "Content-Type: application/json" \
-d '{"api_token":"<your_token>"}'Check iSCSI portal connectivity:
nc -zv <array_iscsi_ip> 3260Common issues:
- API Token Permissions: Ensure the API token has sufficient permissions to create and manage volumes on the Pure Storage array
- Multipath Configuration: Verify multipath.conf is correctly configured and multipath devices are recognized
- Network Connectivity: Check firewall rules and network routes to Pure Storage array
- Plugin Updates: Ensure you are using the latest version of the plugin
LVM inside a volume:
If you plan to use LVM inside a volume, it is better to add purestorage volumes to the ignore list to avoid scanning.
cat /etc/lvm/lvmlocal.conf
...
devices {
global_filter=["r|/dev/zd.*|","r|/dev/rbd.*|",
"r|/dev/mapper/3624a9370.*|"]
}Debug output contamination (#56):
When debug logging is enabled, debug messages may contaminate command
outputs that should be clean (e.g., qm showcmd, pvesm path). This can
break tools that parse these outputs.
Workaround: Disable debug logging when using commands that need clean output:
pvesm set <storage_id> --debug 0Or use temporary debug only when needed:
PURESTORAGE_DEBUG=1 <command>Volume auto-mount issues (#59):
The filesystem_path function returns device paths that may not work
reliably when volumes are deactivated. This affects operations like
qm showcmd and backup tools (e.g., Veeam) that need direct access to
volume paths.
Current behavior: Volumes are activated on-demand and may deactivate when not in use.
Workaround: Ensure volumes are activated before accessing them directly. For automated workflows, consider implementing volume activation in your scripts.
Note: A proper solution using autofs for automatic volume mounting is being evaluated.
Contributions to this project are welcome.