ProblemPossible data corruption when a failed path of a multipathed sliced disk is recovered in Veritas Volume Manager 5.0 on a Solaris environment
SolutionThis issue only applies to configurations where all of the following applies:
1. Veritas Volume Manger (VxVM) 5.0 is installed without a maintenance patch (this issue has been fixed in VxVM 5.0 Maintenance Patch 1).
2. VxVM is running on a Solaris platform (platforms other than Solaris do not support sliced disks).
3. The disks are multipathed.
4. The disks are initialized as sliced disks. Sliced disks have slice 3 created for Private Region and slice 4 created for Public Region.
5. An I/O to the disk fails, but a SCSI inquiry to the disk still returns successfully.
The particular path failure scenario in the above condition 5 triggers the data corruption. In this path failure scenario, dynamic multipathing (DMP) performs the I/O retries with the SCSI driver bypassed. Due to the issue described in Etrack Incident 637898, DMP doesn't take the disk offsets of the public and private regions into account and data from the wrong disk location is accessed and returned back to the application.
Recommended Course of Action
If sliced disks are used with VxVM 5.0 on Solaris, VxVM 5.0 MP1 must be installed to prevent this data corruption. VxVM 5.0 MP1 is part of Veritas Storage Foundation and High Availability Solutions 5.0 MP1. Refer to this TechFile for the procedure to download VxVM 5.0MP1:http://support.veritas.com/docs/288505