RE: Raid problems

Here is a cut and paste of an IBM doc on this subject:

IBM Software Technical Document
__________________________________________________________________
/s_dir/slkbase.nsf/1ac66549a21402188625680b0002037e/e4899b4c32c9116f862565c2007cf565?OpenDocument&ExpandSection=1
 - 
_Section1/s_dir/slkbase.nsf/1ac66549a21402188625680b0002037e/e4899b4c32c9116f862565c2007cf565?OpenDocument&ExpandSection=1
 - _Section1Document Information
__________________________________________________________________

Document Title
Why Damage Can Occur on Raid-5 or Mirrored Drives

Document Description

Question: Why would I get damaged objects on my system that has a form of data 
protection running on it? This could be system-level mirroring or drive-level 
RAID-5 protection.

Answer: The reason that damaged objects can occur is that the basic concept 
behind these techniques is to improve system availability (for example, keep 
the system up) and recoverability (for example, prevent a loss of data) when 
disk drive failures occur. If a disk drive fails or has a problem, I/O 
operations are suspended to it and the system continues to run. When you next 
try and access the failed disk drive, with mirroring the data is read from the 
failed drive's mirrored copy and with RAID-5 the data is rebuilt from the 
parity set contained on the other RAID-5 drives.

As an extra benefit to availability and recoverability, they also offer very 
good data integrity, but it is not guaranteed to be 100 percent. The reason for 
this is because the system is reliant on either the hardware or software 
detecting a problem. When hardware sees a problem, the system functions like 
described above. This is how most data integrity is maintained, and with the 
newer hardware with more modern (robust) error detection techniques, this is 
even better.

As an example of how a damaged object could occur, consider the following 
scenario:
1       A disk drive has a problem and it damages an object internally. The 
hardware somehow doesn't detect that this problem has occurred so the disk 
drive still stays available.
2       The system then requests and reads that object onto the system (main 
memory) for processing.
3       If software does not detect the problem with the object on the system, 
that damaged object could be written back out to disk. With mirroring, both 
mirrored disk drives would then have copies of the damaged object and with 
RAID-5, the parity set would be updated with the information in the damaged 
object
The AS/400 Advanced Backup and Recovery Guide states the following:

Mirrored protection does not protect your system from damage to objects that 
occurs for reasons other than a DASD failure. For example, objects may be 
damaged if your system ends abnormally due to a power failure.




__________________________________________________
Kirk Goins
IBM Certified iSeries Technical Solutions Expert
IBM Certified Designing IBM e-Business Solutions
Pacific Information Systems - An IBM Premier Business Partner
503-674-2985           kirkg@pacinfosys.com


-----Original Message-----
From: Art Tostaine, Jr. [mailto:art@link400.com]
Sent: Thursday, August 29, 2002 2:08 PM
To: midrange-l@midrange.com
Cc: Barry Oppenheim
Subject: Raid problems


A customer of mine has those BCC drives that were talked about a couple of 
months ago.  One of them failed and the raid set showed
UNPROTECTED.

BCC sends new drive overnight (they wouldn't replace same day like IBM would).  
They send Decision Data to replace it.  EXCEPT, we
have tons of damaged objects.  How can objects get damaged if the disks are 
protected by RAID.

Customer called IBM support to talk about this and they say occasionally 
damaged objects can happen, even with Raid.

Has anyone else experienced this?

P.S. My customer may chime in here as he wanted the URL on how to join this 
list.

_________________
Art Tostaine, Jr.
CCA, Inc.
Jackson, NJ 08527

_______________________________________________
This is the Midrange Systems Technical Discussion (MIDRANGE-L) mailing list
To post a message email: MIDRANGE-L@midrange.com
To subscribe, unsubscribe, or change list options,
visit: http://lists.midrange.com/cgi-bin/listinfo/midrange-l
or email: MIDRANGE-L-request@midrange.com
Before posting, please take a moment to review the archives
at http://archive.midrange.com/midrange-l.