|
|
|
Servers - Synchronization to prevent RAID
rebuild failure |
Service Hints &
Tips |
Abstract RAID array synchronization - Prevent
RAID rebuild failure
Summary When a hard drive fails and is
replaced in a RAID-1 or RAID-5 array, a Rebuild failure may
occur if there is a sector on one of the remaining working
drives that cannot be read.
RAID-1 and RAID-5 logical
drives must be synchronized immediately after they are created
to assure that the mirror copies of the data (RAID-1) or
parity data stripes (RAID-5) are correct and match the
associated blocks that they are intended to
protect.
RAID synchronization is recommended as a
preventative maintenance procedure to reduce the risk of an
array Rebuild failure. IBM recommends that synchronization be
run weekly to provide a high level of protection. The level of
protection increases as more frequent synchronizations are
performed. To reduce the frequency of synchronizations to once
or twice a month and still maintain a high level of
protection, schedule synchronizations along with other
preventative maintenance procedures like regular tape
backups.
Over time a hard disk may accumulate grown
defects. This is normal. Defects are corrected on accessed
files by RAID and by the disk subsystems. If a grown defect is
encountered when a file is accessed, both the data and parity
information are corrected using the information striped across
all drives. However, if a grown defect appears on an area that
is not accessed (the area is free space, or because the file
is accessed from cache), then synchronization is required to
reconstruct the data. If all drives are functional, the defect
is reallocated and the data is recovered. However, if a drive
has a grown defect, and another drive has failed completely,
then there is not enough information to reconstruct the data
and the Rebuild will fail.
Synchronization can help to
identify potential drive failures because it scrubs every
block on a logical drive. Synchronization forces all data in
the logical drive to be read regardless of whether the data is
part of a valid file or just free space. The RAID adapter
recomputes the parity based on the data it reads and re-writes
the parity for RAID-5 or just rewrites the mirror copy of data
for RAID-1. The RAID adapter reconstructs any data that it
cannot read due to grown defects and re-writes/reassigns the
data so that it can be read without error later. Then, if a
drive fails, the Rebuild will complete.
In addition,
the RAID Adapter Firmware level should be checked and updated
if necessary. The PCI F/W RAID Adapter requires Firmware 2.43
or later. The MicroChannel F/W Streaming RAID Adapter/A
requires Firmware 2.21 or later.
SAS Keywords
: PSY2, PSY2ADPT,
D/T8640, D/T8642, 320, 06H5078, 06H3059, 92F0335, 06H9334,
DDD, DEFUNCT, 520, 720, SERVER, 500 |
Document
ID: RMIE-3A2CXD |
Last
Modified: 1999-09-11 |
Copyright © 2000
IBM Corporation | | |