The SupremeRAID™ driver introduces support for the NVMe DSM deallocate (trim) command on virtual drives, improving the efficiency of managing unused storage space on NVMe SSDs. This feature enables filesystems or applications to issue deallocate commands on virtual drives, which are translated by the driver and sent directly to the SSDs. By allowing the drives to manage deallocated blocks internally, this reduces write amplification, optimizes storage efficiency, and enhances overall performance.
When a discard command is issued to a virtual drive, it triggers a corresponding deallocate command to the underlying NVMe SSDs. The maximum discard request size reported by our driver is 2TB, and the granularity is 4KB. However, the actual discard request size is restricted by the implementation of the filesystem or application.
This feature is automatically enabled on NVMe SSDs that support the deallocate command and guarantee that deallocated blocks return zeros. To ensure the filesystem can take advantage of this capability and issue discard commands when files are deleted, it must be mounted with the 'discard' option, or using 'fstrim' command to trim unused space periodically.
Note: Not supported on the RAID0 and RAID10 with the strip size larger than 4KB.
Most enterprise-grade NVMe drives perform data integrity checks during read operations. When data corruption is detected, the drive reports an Unrecoverable Read Error (URE) to the host. As a RAID solution with built-in redundancy, SupremeRAID™ can use this redundancy to reconstruct the corrupted data and write it back to the drive, resolving the URE.
In this update, we enhance this process by leveraging the NVMe Get LBA Status capability. Rather than waiting for an application to read the affected block, this feature allows us to proactively identify and repair corrupted blocks. By subscribing to asynchronous events and retrieving LBA status, we can detect and correct potential errors early, minimizing disruption to the application.
Please note that Get LBA Status is not supported by all NVMe drives. However, for drives that do support it, SupremeRAID™ automatically enables the feature, ensuring timely repairs and enhanced data integrity.
This improvement reduces the risk of data loss and significantly improves RAID efficiency on NVMe SSDs.
- The management GUI is installed and enabled by default.
- The password for the administrator account can be set up during the installation process.
- A quick setup wizard has been added for license activation and basic configuration.
- The consistency check management page has been refined for better usability.
Users can now configure Self-Encrypting Drive (SED) keys directly when creating physical drives, eliminating the need for third-party tools to lock the drive beforehand. To take ownership of a physical drive with SED support (if the drive is not yet locked), use the --sed-take-ownership option. This command will prompt for confirmation, a new SED key, and any required credentials.
Note: This process will erase all user data on the drive.
$ sudo graidctl create physical_drive /dev/nvme1 --sed-take-ownership
To bypass prompts, you can use the following options:
$ sudo graidctl create physical_drive /dev/nvme1 --sed-take-ownership --new-sed-key [New Password] --no-current-sid --confirm-to-erase
$ sudo graidctl create physical_drive /dev/nvme1 --sed-take-ownership --new-sed-key [New Password] --current-sid [Original Password] --confirm-to-erase
$ sudo graidctl create physical_drive /dev/nvme1 --sed-take-ownership --new-sed-key [New Password] --psid [PSID] --confirm-to-erase
Note: During ownership, both the SID and admin1 key will be set to the same value (the SED key), and only this SED key will be stored in the system.
This version of the driver installer is outdated. Please use the latest stable version for deployment.
The following models have passed Supermicro's validation for drive LED indicator behavior.
- Environments with an NVIDIA NVLink Switch System are currently not supported due to an NVIDIA driver limitation.
- The failed indicator would not indicate when SAS/SATA drives have been removed.
- The drive status will not turn into missing immediately when the SAS/SATA drive is pulled out.
- On some server models, slot IDs for SAS/SATA physical drives might not display properly.
- Did not support NVMeoF physical drives which exported from Ubuntu 20.04 (kernel < 5.15).