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

scsimgr(1M)
The scsimgr command provides a single command line interface to manage and diagnose the mass storage
stack. It is designed to work with persistent DSFs, but also works with legacy DSFs for a limited set of operations.
The scsimgr command includes the following features:
• Retrieves and clears driver statistics
• Displays status information about SCSI objects
• Retrieves, sets, or saves attributes of SCSI objects
• Disables and enables SCSI objects
• Performs SCSI task management functions such as LUN and target resets
• Performs miscellaneous SCSI commands such as inquiry and self-tests
In addition to gathering statistics and printing device information, you can use scsimgr to set attributes, which
replace SCSI tunable parameters set at the operating system level. Attributes can be set globally like kernel
tunable parameters, or can be restricted to a particular device type, device instance, driver, SCSI target, vendor,
or product.
In HP-UX 11i v3, scsimgr provides generic management capabilities for the SCSI subsystem and driver-specific
management capabilities for disk drivers.
For more information, including a list of commands and keywords for the scsimgr command, see the
scsimgr(1M) man page and Appendix C: Using scsimgr
. A more detailed white paper entitled Scsimgr SCSI
Management and Diagnostics Utility is available on the web; see For more
information.
Obsolete and Deprecated Features
Infinite I/O Retries
In previous releases, certain types of disk I/O request failures were retried indefinitely by the mass storage stack.
Starting with HP-UX 11i v3, this behavior is configurable; the default behavior retries failing I/O requests a finite
number of times. If all the retries fail, the I/O request returns a failure notification to the calling application.
Some applications are designed to expect I/O requests to always succeed. HP recommends testing your
applications with the new finite retry policy to determine how they behave when an I/O request fails.
To control the retry policy, use the scsimgr command to set the infinite_retries_enable (which toggles
the policy between infinite and finite) and max_retries (which defines the number of retries in the finite case)
attributes. Changes to these attributes take effect immediately and do not require a reboot.
For example, to restore infinite retries, enter the following command:
# scsimgr set_attr –a infinite_retries_enable=true
To set the retry policy for a particular device to finite retries, enter the following:
# scsimgr set_attr -d esdisk -D device_file –a infinite_retries_enable=false
To change the number of retries for all disks, enter the following:
# scsimgr set_attr -N /escsi/esdisk -a max_retries=new_value
Note: To set the number of retries, the retry policy must be set to finite.
25