Skip to main content

Overview

Polystack Block Storage supports both thin and thick provisioning models. Thin provisioning allocates storage capacity on demand — physical space is consumed only as data is written, not when the volume is created. Thick provisioning reserves the full volume capacity immediately at creation time. The provisioning behavior is controlled by the backend driver and the scheduler overcommit configuration.
Administrator Access Required — This operation requires the admin role. Contact your Polystack administrator if you do not have sufficient permissions.

Thin vs Thick Comparison


Backend Provisioning Support

Ceph RBD images are sparse by design. When a volume is created, no physical storage is consumed until data is written. No configuration is required to enable thin provisioning on the Ceph backend.
cinder.conf — Ceph RBD backend
Because Ceph RBD is thin by default, the max_over_subscription_ratio setting in the [DEFAULT] section controls how aggressively the scheduler overcommits capacity. The default value is 20.0 (20x overcommit).
LVM uses thick provisioning by default. To enable thin provisioning, create a thin pool in the volume group before configuring Cinder, then set lvm_type = thin in the backend section.Step 1: Create the thin pool
Create LVM thin pool
Step 2: Configure Cinder
cinder.conf — LVM thin backend
Step 3: Apply the configuration
Deploy block storage configuration
Step 4: Verify thin volumes are created
Check LVM thin volume allocation
Volumes created from this backend will show Twi in the Attr column, confirming thin layout.
NetApp ONTAP supports thin provisioning through FlexVol volumes. Set netapp_lun_space_reservation = disabled to enable thin provisioning on the LUN.
cinder.conf — NetApp thin provisioning
Pure Storage FlashArray volumes are thin-provisioned by default. The array reports the provisioned size to the scheduler, not physical consumption. No additional configuration is needed.
cinder.conf — Pure Storage thin provisioning

Scheduler Overcommit Configuration

The Cinder scheduler uses max_over_subscription_ratio to determine how much capacity to advertise to tenants above physical availability. This applies to backends that report thin_provisioning_support = True.
cinder.conf — scheduler overcommit settings
Setting max_over_subscription_ratio too high without monitoring physical usage can result in volumes that cannot be written to once the physical backend is full. Monitor actual consumption against provisioned capacity and set alerts at 70–80% physical utilization.

Monitor Overcommit and Capacity

Navigate to Storage > Storage Backends (admin view) to view backend capacity statistics. The Free Capacity and Total Capacity columns show the physical pool state reported by each backend driver.For provisioned-vs-physical overcommit visibility, use the Monitoring dashboard to track Cinder backend pool utilization over time.

Next Steps

External Storage Integration

Connect to enterprise storage backends including NetApp, Pure Storage, and Dell

Storage Backends

Configure and register backend drivers with the scheduler

Volume Types Admin

Expose provisioning mode to tenants through volume type extra specs

Storage Tiers

Configure multi-tier storage with NVMe, SSD, and HDD backends