Notification
Risk classification
HIPER (High Impact and/or Pervasive)
Risk categories
Data Loss
Abstract
In IBM Storage Scale versions 5.1.0.0 through 5.2.2.1 (IBM Storage Scale System 6.1.0.0 through 6.2.2.1), IBM has identified a potential integrity issue for file system data. Under certain conditions, incorrect snapshot data —either stale or uninitialized— might be read while the mmchdisk start command is being executed on file systems with replication enabled.
Description
The mmchdisk start command is used to bring disks in the down state back to the up state. In replicated file systems, this command also attempts to repair stale data on these disks. If a replica on a down disk is not written during a file block update, that replica may become stale or remain uninitialized until the completion of the mmchdisk start command, which fixes the situation. Stale replicas are not read by applications.
After the mmchdisk start is issued, the affected disks enter an unrecovered state. Upon successful completion of the command, these disks transition back to the up state, and all replicas become readable by applications again.
Tip: The mmlsdisk command can be used to check disk states. Any disk in the unrecovered state gets listed in the output.
If a workload accesses the snapshot data while a disk remains in the unrecovered state, a problem has been uncovered where there is a risk of reading stale or uninitialized data —if the affected disks have not yet been repaired.
Problem determination
A generic "file system struct error" may occur, although this message is not specific to the issue described here. It is also possible for incorrect data to be read silently, without any overt error symptoms.
If such an error is triggered, the /var/log/messages file (or the output of the errpt command on AIX) might contain an entry similar to the following:
Error=MMFS_FSSTRUCT, ID=0x94B1F045, Tag=12662454: Invalid disk data structure. Error code 1108.
Note: The MMFS_FSSTRUCT message is general and may not directly indicate this specific problem.
Users affected
Customers that run IBM Storage Scale versions 5.1.0.0 through 5.2.2.1 (IBM Storage Scale 6.1.0.0 through 6.2.2.1), particularly if they are accessing snapshot data while the mmchdisk start command is in progress.
This issue may occur when replication is enabled for user data, metadata, or both.
Recommended Action
To avoid this issue, take the following action:
- Upgrade all nodes to IBM Storage Scale version 5.2.3.0 (IBM Storage Scale System 6.2.3.0) or later:
Until the fix can be applied
While the affected customers should upgrade their systems at the earliest opportunity, refer to the following instructions to reduce the chance of data loss.
Avoid accessing snapshot data while running the mmchdisk start command.
If an upgrade is not possible, customers should contact IBM Support and reference APAR IJ54328.
Reference ID
Internal reference: D.339231
Date first published
08 July 2025
Was this topic helpful?
Document Information
Modified date:
08 July 2025
UID
ibm17232143