IBM FlashSystem A9000/R: In extremely rare cases, in any type of mirroring relationship: synchronous mirroring, asynchronous mirroring, HyperSwap, or Multi-site, a bug in the microcode might lead to undetected data corruption on the Primary system

In extremely rare cases in any type of mirroring relationship: synchronous mirroring, asynchronous mirroring, HyperSwap, or Multi-site, a bug in the microcode might lead to undetected data corruption on the Primary system. After the below certain sequence of events and in a specific timing,  invalid data may be returned to the host, as undetected data corruption.

Affected versions:
A9000/R systems running versions 12.3.2 to 12.3.2.b

Fix:
12.3.2.c - released on Jan 18, 2021

Reference: SYS-316182

The following complete sequence of events need to happen to hit this issue:

1. Mirror configured between Primary (A) and Secondary (B) systems
2. Mirror is being deactivated - User action
3. User writes zeroes to partition on the Primary (A) system
4. Change role - Primary (A) to Secondary (B) - User action (change role can be done only upon deactivating the mirror relations)
5. Mirror is being reactivated - User action
          a. In order to hit the issue, while doing this action an internal race condition needs to happen and within an interval time of 1 second which is extremely rare 
6. Change role - Primary (B) to Secondary (B) - User action 
   • If 5 and 5a conditions above were met the partition on the current Primary (A) will not be resynced

Probability: Extremely rare

Mitigation and IBM Recommendation:

• The issue can be avoided if any one of the above events is eliminated
• Upgrade to version 12.3.2.c as soon as possible
• If you plan to perform the above sequence of events and you can't upgrade at this time, contact your IBM Technical Advisor should you have one assigned, or your IBM Representative or IBM Support - to discuss alternatives