IBM ServeRAID-8e may incorrectly handle hard drive medium errors - Servers

Resolving The Problem

The system may be any of the following IBM eServer xSeries:

• an xSeries 306m (Type 8891), any model
  
• an xSeries 306m (Type 8849), any model
  
• an xSeries 206m (Type 8485), any model
  
• an xSeries 206m (Type 8490), any model
  

The system may be any of the following IBM IntelliStations:

• an IntelliStation M Pro (Type 6218), any model
  

The system has the symptom described above.

The fixes are provided in ServeRAID BIOS and device drivers specified below:

• SAS BIOS level 1554 (This is a separate download flash update)
  
  
• SATA BIOS level 1217 (This flash update is integrated into the System BIOS update)
  
  
• ServeRAID-8e (SAS Controller Drivers)
  Linux Driver : version 1.1.4425.818
  Windows Driver : version 1.1.4425.818
  NetWare Driver : version 1.1.4425.818
  
  
• ServeRAID-8e (SATA ICH7R Drivers)
  Linux Driver : version 1.19.3854.V041A8
  Windows Driver : version 1.0.3854.41
  

These files can be downloaded from the IBM ServeRAID Software Matrix.

None.

The ServeRAID-8e uses two software components to handle error recovery like a medium error detected by a hard disk drive. The two components are the ServeRAID-8e BIOS and the operating system device driver. The BIOS code handles this type of error recovery throughout the system's boot process, until the operating system's (OS) device driver is initialized and assumes this role. The BIOS code also handles all int13 calls, which is typically what is used by system imaging tools for the
purposes of unattended OS installations. After the OS is running, the ServeRAID-8e device driver handles all disk related error recovery.

IBM has determined that hard disk drive medium errors were not being correctly handled by both of these components. An incorrectly handled medium error while operating in a redundant RAID-1 mirrored configuration typically does not result in immediately visible symptoms to the user as the RAID-1 redundancy provides another path to read or write the data to the mirrored pair. Typically the problem occurs later when one of the drives in the mirror is marked defunct and the surviving drive retains an uncorrected medium error. When this happens, there is no way to recover the missing data in that location. There is also a very remote possibility that incorrect data may be read from that location.

The updated BIOS and Device Drivers improves the medium error handling, mirror integrity, and any incorrect events will be reported to the user via the IBM provided software package, ServeRAID Manager. IBM strongly recommends that all customers with affected models upgrade to the new BIOS and device driver as soon as possible to avoid potential data integrity issues.