Overview: The Next Generation Mass Storage Stack (September 2009)

binding, the DSF name and minor number no longer have hardware path information encoded in them. Instead,
they have a simple device instance number such as /dev/disk/disk3. Rather than mapping to the LUN
hardware path, the persistent DSF maps to the LUN WorldWide Identifier (WWID), a resource on the LUN similar
to a serial number. The WWID is unique to a LUN and does not change for that LUN. If any of a LUN lunpaths
change, the DSF bound to the LUN does not change – hence, the DSF has agile addressing.
Impact
HP encourages the use of persistent DSFs; they are created by default on newly installed servers. Existing DSFs
continue to work as before, and are retained on an update to HP-UX 11i v3. Legacy DSFs are backward
compatible and are not affected by persistent DSFs on the same server. A device can be simultaneously accessed
via legacy and persistent DSFs.
All commands are backward compatible, and work with either legacy or persistent DSFs, except as noted in
Backward Compatibility
. Some commands have new options to manage persistent DSFs and DSF migration:
insf
By default, creates both persistent and legacy DSFs for new devices.
insf -L
Restores legacy DSFs and legacy configuration information.
rmsf -L
Aids in migration by removing all legacy DSFs and legacy configuration information.
ioscan –m dsf
Maps persistent DSFs to their equivalent legacy DSFs and vice versa.
ioscan -N
Prints persistent DSFs when used with the –n option to list DSFs.
ioinit –rC
class
Reassigns device instance numbers, starting from 0 (zero), for all existing devices in the
given class. To create new DSFs for the devices whose instance number has changed, run
insf –e -C class.
ioinit –A
Performs Critical Resource Analysis (CRA) on all of the devices corresponding to the
hardware paths specified in infile, before assigning the new instance numbers. You
must use the –f option with this option.
io_redirect_dsf
Associates a new disk with an existing set of DSFs. This is used when replacing an internal
disk or a disk in a JBOD.
For additional information on the mass storage device naming conventions in HP-UX 11i v3, see the HP-UX 11i v3
Mass Storage Device Naming white paper in For more
information.
Multi-Pathing and Load Balancing
Background
Agile addressing creates a single DSF for each mass storage device, regardless of the number of hardware paths
to the disk. The mass storage stack in HP-UX 11i v3 uses that agility to provide transparent multi-pathing. In other
words, if a LUN has multiple lunpaths, I/O requests can be transparently distributed across all available lunpaths
to the LUN, using a choice of load balancing algorithms. This eliminates the need for add-on multi-pathing
products.
If a lunpath fails, the mass storage stack automatically disables the failed lunpath and I/O continues on all
available enabled lunpaths. Any failed or nonresponsive lunpaths are monitored, so that when a failed lunpath is
recovered it is automatically and transparently reincorporated into any load balancing. In addition, any new LUNs
or new lunpaths are also automatically discovered and added to load balancing.
Impact
If none of your disks are multi-pathed, there is no impact to your server.
Existing legacy DSFs automatically inherit multi-pathing. I/O requests and I/O control operations can be
processed along any hardware path to a LUN, even if you use legacy DSFs. If you do not want multi-pathing of
legacy DSFs, disable it using the new scsimgr command. For more information, see Disabling Multi-Pathing on
Legacy DSFs.
6