User`s guide

• Your RAIDsets can achieve the high performance characteristics of RAID
level 3 provided you set your RAIDset chunksize to the minimum value (16)
and your application calls for large sequential I/O operations.
• RAIDsets and mirrorsets can contain disks of different sizes, but the disk
space used by the storageset is limited to the maximum size of the smallest
disk in the RAIDset.
• Place RAIDset and mirrorset members on different ports. This keeps your
storageset from going inoperative in the event that a single port bus failure
occurs.
• You may have a combined maximum total of 20 mirrorsets and RAIDsets.
• You may have a combined maximum total of 30 storagesets (mirrorsets,
RAIDsets, stripesets).
• You can have no more than 32 physical device members for a unit (practically
this applies to RAID 0 & 1 configurations).
6.4 Avoiding Unwanted Unwritten Cached Data Conditions
Write-back cache is required for RAID functionality. Therefore, you may
experience situations that leave unwanted unwritten cache data in the write-back
cache module. This section should help remedy some of those situations.
Unwritten cached data can remain in cache for reasons other than a power
failure. To avoid some of these situations when write-back cache is enabled,
follow these recommendations:
• When configuring dual-redundant controllers for failover, make sure the
target controller (the controller you are copying configuration information
‘‘to’’) has no cache errors or unwritten cached data. Delete any units on the
target controller to verify there are no cache errors, before entering the SET
FAILOVER command.
• Do not take a dual-redundant controller pair out of failover (by entering
the SET NOFAILOVER command) with unwritten cached data present
in the write-back cache modules. Doing so will destroy data. Enter the
SHOW THIS_CONTROLLER and SHOW OTHER_CONTROLLER commands
to confirm that cached data has been written.
Note
When the SET NOFAILOVER command is issued, the other controller
shuts down.
• When write-back caching is enabled on a unit and that unit fails, the cached
data for that unit is unwriteable. The data for other write-back cached units
on the same controller are still intact and the write-back cache module is
still fully functional. However, in order to clear that portion of the write-back
cache module’s memory that now contains data from the failed unit, you need
to enter the CLEAR_ERRORS UNWRITEABLE_DATA command.
• In order to force a cache data flush of an online unit, enter the
SET unit NOWRITEBACK_CACHE command. This forces a flush of any
outstanding write data (completes within several minutes).
6–8 Working with RAID Arrays