HP StorageWorks HSG60 and HSG80 Array Controller and Array Controller Software Command Line Interface Reference Guide (EK-G80CL-RA. C01, March 2005)

CLI Command Descriptions
190 HP StorageWorks HSG60 and HSG80 Array Controller and Array Controller Software
Command Line Interface Reference Guide
is only to be issued through one controller. The controller must contain a valid
cache configuration before this switch is assigned. Unlike going from
Nonmirrored mode to Mirrored mode, going from Mirrored mode to
Nonmirrored mode is permitted with a failed cache module. The controller
automatically restarts if this switch is specified.
Modifying the MIRRORED_CACHE setting while the system is running
requires stringent attention. If the subsystem is connected to hosts that are
running, you must carefully observe and monitor the CLI output directly after
invoking this command.
After changing the operational mode for cache (MIRRORED_CACHE or
NOMIRRORED_CACHE), a warning message is displayed:
Top>SET THIS NOMIRROR
Warning: Changing mirroring mode causes persistent
reservations to be lost. If there are units listed as having
persistent reservations in the following message, then
please use the command SET UNIT FAKE_PR so the host can clean
up the persistent reservations.
Unit D20 has a persistent reservation.
Unit D65 has a persistent reservation.
Unit D66 has a persistent reservation.
Unit D103 has a persistent reservation.
A list of units may be displayed that identifies which units, if any, have an
active persistent reservation set within the cache. After this output is
complete, the change in the mirrored cache configuration is made.
Since Tru64 UNIX host systems automatically assume that persistent
reservations are never lost and array controllers automatically restructure
cache data whenever its operational mode is changed, additional steps must be
taken. First, you must associate any one unit with persistent reservations with
each cluster. Then, you must enter the SET unit FAKE_PR command to
restore persistent reservations that may have been lost. Lastly, you must
execute any type of Tru64 UNIX host operation that results in a read to the
physical unit associated to restore all persistent reservations. Doing this,
causes Tru64 UNIX host systems to re-establish unit persistent reservation
structures.
If, after examining displayed units, you discover that more than one node or
cluster of nodes has units with persistent reservations on a unit, you must
invoke the FAKE_PR switch against a unit for each different cluster or node.