Certain 3.84TB NVMe drives on IBM FlashSystem or Storwize systems (identified below) may result in data loss or undetected data corruption

This exposure relates to 3.84TB NVMe drives, with product ID 1014063B and firmware SN1E only. These drives can be used in FS9110, FS9150, V7000 Gen 3 and V5100 systems. The issue does not affect IBM FlashCore Modules, or any drives in SAS enclosures.

During normal operations, a table with metadata used for establishing an encrypted connection to the drive can get written incorrectly. While no symptoms are apparent when the bad data is written, operations which require a reconnection to a drive can result in a drive failure.  It is possible for this to occur to multiple drives leading to a loss of access, or in some cases loss of data. When the incorrect metadata is written, it can also cause data on some of the drive to be cleared, which is detected when the data is read on systems with Data Reduction Pool (DRP). On systems without DRP, the cleared data may not be detected. The problem can be detected on Storwize products via a background process that validates whether the data on disk matches the parity on disk. If the background process detects that the data does not match the parity, for example when this drive issue occurs, then the system will log a 1691 error. However, the data could be read prior to the background validation.

IBM strongly recommends that customers with the specified drive models do not perform any maintenance action and contact IBM Support. These maintenance actions include:

IBM also strongly recommends that any affected systems which are not yet in use should not be moved into production, until the drives are running a new version of firmware.

In all instances, customers with potentially affected drives (3.84TB NVMe drives (with product ID 1014063B and firmware SN1E or SN1ESN1E) on the systems listed above should contact IBM Support immediately to evaluate exposure to this issue and obtain a fix in a new drive code level.