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

Table Of Contents
# ioscan –r -H device_hw_path
Disabling Multi-Pathing on Legacy DSFs
By default, the multi-pathing feature of the next generation mass storage stack enables I/O requests and I/O
control operations to be processed along any hardware path to a LUN. Even if legacy DSFs are used for I/O,
requests can still be routed through a different hardware path. This maximizes availability and parallelism.
To force legacy DSFs to use backward-compatible multi-pathing behavior, you can use the scsimgr command to
configure a global device tunable called leg_mpath_enable. If you set leg_mpath_enable, multi-pathing is
enabled on all legacy DSFs, and I/O control operations (ioctls) are processed on any available path. This is
the default behavior. If leg_mpath_enable is not set, no multi-pathing is performed on any legacy DSF, and
ioctls are processed only on the LUN path corresponding to the legacy DSF. To force the backward-compatible
behavior persistently for the entire server, enter the following:
# scsimgr save_attr -a leg_mpath_enable=false
To limit the behavior to a particular LUN, set the per-LUN device tunable, which overrides the global value as
follows:
# scsimgr save_attr -D /dev/rdisk/disk4 -a leg_mpath_enable=false
Note: Multi-pathing through persistent DSFs is not affected by this tunable.
Changes to Other Subsystems
This section describes how other parts of HP-UX are affected by the next generation mass storage stack. Most
subsystems that deal with I/O or mass storage have been updated to use the agile view. They allow the use of
both persistent and legacy device special files, and the use of legacy, lunpath, and LUN hardware paths. Some
subsystems only support one of the two views, and some commands have new options related to the mass storage
stack; these are summarized below.
Crashdump
The dump subsystem supports both the legacy view and the agile view. If a dump device is configured using a
legacy DSF, crashconf converts it to an equivalent persistent DSF and selects an available lunpath hardware
path for the dump device.
The dump subsystem is aware of multi-pathed devices and supports automatic dump device path failover. If a
configured path goes offline, the dump subsystem automatically selects an alternate available hardware path and
reconfigures the dump device.
When a LUN dynamically expands or contracts, the dump subsystem automatically updates its internal data
structures to use the new device size.
Using lvlnboot, vxvmboot, and the /stand/system file to configure dump devices is deprecated and will be
obsoleted in a future release.
For additional information on the dump subsystem in HP-UX 11i v3, see the HP-UX 11i v3 Crash Dump
Improvements white paper in For more
information.
28