Current Series Release Notes

29.0.0.0rc1-61

New Features

  • NetApp – The new ASAr2 driver now supports the snapshot and cache acceleration workflow. It leverages NetApp snapshot technology to create, delete, and restore snapshots, as well as to create clone and volumes from snapshots.

    The following features are supported by the ASAr2 driver:

    1. Snapshot Creation.

    2. Snapshot Deletion.

    3. Snapshot Revert.

    4. Create Volume from Snapshot.

    5. Creating Volume from Source Volume.

    6. Create Volume from Image.

  • Dell PowerStore driver: Added support for manage and unmanage existing volumes.

  • Infinidat driver: Added support for snapshot promotion, available in InfuzeOS 8.1 and later. Creating volumes from existing volumes or snapshots now uses snapshot promotion when supported by the storage system, producing independent volumes without copying data. Systems without this capability continue to use generic data copy.

    Promoted volumes follow the same provisioning and compression settings as newly created volumes. The san_thin_provision option selects thin or thick provisioning, and infinidat_use_compression controls compression. When infinidat_use_compression is unset, the configured pool’s compression setting is applied. Volumes are extended when the requested size exceeds the size inherited from the source.

  • Adds a new glance_preferred_stores configuration option. The option is an ordered list of Glance store identifiers that Cinder passes to Glance when downloading an image, allowing deployments with replicated images to prefer a local or co-located store.

    This feature requires Glance 2026.2 (Hibiscus) or later with multi-store enabled and python-glanceclient 4.12.0 or later. On an older Glance deployment, the prefer query parameter is silently ignored and the option has no effect.

    Troubleshooting:

    • Glance rejects the request with HTTP 400 if multi-store is not enabled or if a configured store is not in its enabled_backends. A typo in this list will therefore cause all volume creation from image requests to fail.

    • Glance does not apply the preference when its image cache middleware serves the download, so the option will have no effect in that case.

    • Direct file downloads, when allowed_direct_url_schemes is configured, bypass the Glance client and are unaffected by this option.

  • NetApp ONTAP driver: Added pool-level capability netapp_san_active_active that exposes the ONTAP 9.19.1 SVM-level san_multipathing property (active_active or local_active) as a schedulable extra spec. Operators can steer volume placement toward Active-Active or Local-Active backends via volume types using netapp_san_active_active=<is> True or netapp_san_active_active=<is> False. The ZAPI client uses a lazy, cached RestNaServer to query this REST-only property without requiring migration to the full REST client. The REST client queries it natively. The capability is injected into each pool’s SSC data during update_ssc() and excluded from ASAr2 backends where Active-Active is inherent. A version gate ensures the capability is only advertised on clusters running ONTAP 9.19.1 or later; older releases omit the key entirely.

  • NetApp ONTAP driver: Added a new boolean configuration option netapp_use_metrics_based_utilization (default True). When enabled, the driver forces use of the REST client (regardless of netapp_use_legacy_client) and derives the pool utilization value from the ONTAP REST node metrics endpoint instead of the legacy counter-based calculation, reducing ONTAP management API load. ONTAP 9.8 or later is required. This REST override is intentional because the node metrics API is REST-only; operators who must keep the ZAPI/legacy client should set netapp_use_metrics_based_utilization = False. ASA r2 (disaggregated) backends always use the REST client regardless of this option, but this option still controls their utilization value: True reports the metrics-based value averaged across cluster nodes, False always reports the default utilization value since ASA r2 has no counter-based fallback.

  • Volume usage notifications now include cluster_name in addition to host.

Upgrade Notes

  • Infinidat driver: The minimum required InfiniSDK version is now 250.0.0. Install or upgrade infinisdk before restarting the cinder-volume service. The driver will not initialize if the library is missing or older than the required version. This requirement also applies to systems that use generic data copy.

  • The cinder-manage db sync command for this version of cinder will add an additional database index backups_volume_id_data_timestamp_idx on the backups table. Depending on the table size, this will take time to complete.

Security Issues

  • This release contains a fix for CVE-2026-93856, Cross-project information disclosure in the Cinder Messages API. See Bug #2160671 for details.

  • This release contains a fix for Bug #2163356 whereby the cinder ssh injection sanitizer utility, could be tricked into allowing a disallowed character through unescaped in a command from a cinder driver to the backend storage, leading to possible remote code execution on the backend. This bug affected only those cinder drivers that use ssh to communicate with the storage backend and use the sanitizer utility, which include IBM Storwize/SVC, Inspur InStorage, TOYOU ACS5000, the generic SAN driver base class, and the Brocade/Cisco FC zone managers.

  • Bug #2163562: Cinder now blocks invalid iSCSI IQNs when creating attachments with the LIO iSCSI target driver. This target driver is typically used with the LVM volume driver. This prevents manipulation of files/paths on the cinder-volume host.

Bug Fixes

  • Bug #2160671: Fixed a cross-project information disclosure issue in the Block Storage API v3.3 (CVE-2026-93856), where a user with the project ‘member’ or ‘reader’ role making the “List messages” call could find the non-expired user messages for other projects included in the response.

  • Bug #2163356: Improved the ssh injection sanitizer utility, which could be tricked into allowing a disallowed character through unescaped in a command from a cinder volume driver to the backend storage, as long as the first occurrence of the character was escaped, leading to possible remote code execution on the backend. This bug affected only those cinder drivers that use ssh to communicate with the storage backend and use the sanitizer utility, which include IBM Storwize/SVC, Inspur InStorage, TOYOU ACS5000, the generic SAN driver base class, and the Brocade/Cisco FC zone managers.

  • IBM Spectrum Virtualize family driver: Bug #2167561 https://bugs.launchpad.net/cinder/+bug/2167561 Fixed an issue where driver initialization unconditionally executed iSCSI IP discovery (``lsip`) even when Fibre Channel (FC) was configured, which caused setup failure on backends without configured iSCSI IPs.

  • Bug #1696715: Optimize getting parent backup for new incremental by not fetching all backups from the database, but using a DB query to find the suitable parent. Alongside, a previous bug with finding parents for snapshot backups was fixed. Previously this could select an ineligible parent when backing up a snapshot older than existing backups; that case now raises an error instead of silently creating an incremental against the wrong parent. Additionally the backup status of RESTORING was added to the list of valid status, as a restoring backup is indeed a valid, read-only source for a new incremental backup.

  • Dell PowerFlex driver: Fixed image cache vTree size check to count only direct child volumes of the cache volume. Previously, the driver used query_vtree_statistics, which counted all descendants in the vTree, including snapshots and deeper nested volumes. This could incorrectly block clone creation before the cache volume reached the configured direct-child limit. The driver now uses query_vtree_volumes with the ancestorVolumeId filter to count only immediate children. The powerflex_max_image_cache_vtree_size configuration option name and behavior are preserved.

  • Bug #2173158: Fixed the attachment access check added for bug #2138566, which let a non-admin user show an attachment of another project, including its connection info, when their own project had any volume and they knew the attachment’s UUID. The check now looks at the project of the volume the attachment belongs to.

  • [Everpure] Fixed an issue where replicated volume types could be added to a non-replicated consistency group without error.

  • Everpure driver: The backend is now reported to the scheduler as a single pool named after volume_backend_name. The scheduler only emits capacity.pool and capacity.backend notifications for backends that report a pools list, so capacity telemetry had never been produced for FlashArray backends. Existing host#pool values and capabilities.<metric> references are unchanged.

  • Bug #2160366: Fixed an issue in the Pure Storage FlashArray driver where several operations could silently fail when an array limit (such as the maximum volume or snapshot count) had been reached. The py-pure-client REST response was not checked, so the operation was reported to Cinder as successful even though the array did nothing. The driver now inspects the response and fails with a clear error. Affected operations include volume create, volume create from snapshot, cloned volume create, volume extend, revert to snapshot, snapshot create, consistency group create, consistency group snapshot create, consistency group clone and replication failover.

  • Bug #2160366: Fixed the Pure Storage FlashArray driver revert to snapshot operation, which passed the snapshot and volume names in the wrong order and so was silently rejected by the array (the error was previously ignored). Revert to snapshot now correctly overwrites the volume with the snapshot contents.