Disk Surprise Removals or Hyper-V fail overs with SDDDSM

Content

There is an issue in SDDDSM 2480 when MPIO performs two or more path fail over (GmInvalidatePath ) calls for a disk device to SDDDSM around the same time requiring synchronization.
This causes some inconsistent device/paths states to be maintained between MPIO and SDDDSM, which eventually causes MPIO to unexpectedly remove the disk devices.

The issue can be identified with Event ID 157 "Disk # has been surprise removed" in the system event log or the following sequence of traces in the sdd_log file with activeDevCnt going to 0 for all disk devices affected.

Example:

03/07 19:10:00 GmInvalidatePath(206) DSM_FATAL_ERROR failPathid 151c8a90 fogstate 2, bestPathid 151da920 is selected fogstate 1

03/07 19:10:01 GmInvalidatePath(206) DSM_FATAL_ERROR failPathid 151da920 fogstate 2, bestPathid 151e9980 is selected fogstate 1

GmRemovePathFromFailoverGrp(5128) Path 15205b90-6005076801820034280000000000010B-0001000700 removed from FOG 151e9980-00010007.

GmSetPathState(4170) MPD-000-6005076801820034280000000000010B, activePathCnt 0, Path-7, oldState 8, newState 0, FOG 151e9980-00010007, activeDevCnt 0

Recommended Fix

SVC/Storwize customers, we recommend installing version level 2.5.1.0.
SDDDSM 2.5.1.0 does not support DS8000 storage - Check SSIC for supported Multi-path Drivers
http://www-03.ibm.com/systems/support/storage/ssic/interoperability.wss
Customers using DS8K storage should consult their IBM Technical Advisor