scsiadminswap(1M)scsiadminswap(1M)NAMEscsiadminswap - perform a live plug or unplug of a scsi device
SYNOPSISscsiadminswap [ options ]
DESCRIPTIONscsiadminswap allows a user to live plug or unplug a scsi device assuming
it is not a primary resource (such as the system disk). There are three
required options. These indicate whether to perform a plug or an unplug
and provide a bus and device number. An additional argument can be used
to change the default timeout value. To execute scsiadminswap, a user
must have root status.
There are many different ways that a disk may be accessed. While this
man page outlines and gives examples of the more common ones, if there is
any question the user should check the man pages for umount, mount, or
the Insight book "SGI Admin: IRIX Admin: Disks and Filesystems." If the
disk is being used as part of an xlv volume more information can be found
via man on xlv_mgr, xlv_make, xlv_assemble, xlv_shutdown, or chapter 4 of
the above Insight book.
Unplugging a disk:
Before executing the scsiadminswap unplug, all the file systems of the
disk in question must be unmounted (see umount(1M)), detached (if a
mirrored xlv volume) see xlv_mgr(1M). To do so it may be necessary to
use fuser(1M) to kill any processes using the file systems. If a file
system on the disk is an unmirrored xlv volume it is also necessary to
execute an xlv_shutdown(1M) after the umount. For each file system on
the disk one of the following steps should be taken. After one of the
below cases (or other as needed) has been performed for all file systems
on the disk then run scsiadminswap-u -b # -d #.
Case 1: the file system is not an xlv volume
1) Unmount the file system
a) Hint: fuser finds and kills processes using a file system.
Case 2: the file system is an unmirrored xlv volume:
1) Unmount the file system
a) Hint: fuser finds and kills processes using a file system.
2) Perform an xlv_shutdown.
Case 3: the file system is a mirrored xlv volume:
1) Run xlv_mgr and perform a detach of the plex using the disk.
Plugging in a disk:
First run scsiadminswap-p -b # -d #. Now the file system may need to be
mounted (see mount(1M)) or made part of an xlv volume. If the disk is
going to be part of a mirrored volume and already has vlx header
Page 1
scsiadminswap(1M)scsiadminswap(1M)
information, an xlv_assemble(1M) should be performed. If the disk is
going to be part of a mirrored volume but does not have xlv header
information an xlv_make(1M) should be performed before the
xlv_assemble(1M). If the disk had old (from a different volume) xlv
information then perform a delete using xlv_mgr(1M) before either of the
above steps). Depending on what needs to be done one of the following
cases may apply.
Case 1: the file system is not an xlv volume
1) Mount the file system
Case 2: the file system contains no prior xlv information
and will be an unmirrored xlv volume:
1) Perform an xlv_make
2) Run xlv_mgr and perform an attach.
3) Mount the file system
Case 3: the file system is on a disk without prior xlv
information and will be a mirrored xlv volume:
1) Mount the file system
2) Perform an xlv_assemble.
Case 4 & 5: As 2 & 3, except the disk has old xlv_info:
1) Run xlv_mgr and perform a delete of the xlv label information
2) Go to Case 2 or 3
The above cases for plug and unplug only cover the common xlv situations.
More information regarding xlv volumes can be found in the "SGI Admin:
IRIX Admin: Disks and Filesystems" Insight Book. Also see the .B
Examples section at the end of the man page for a sample session.
The required options are:
-u or -p Indicates whether operation is an (u)nplug or a (p)lug. If
both are specified, it will be report as an error.
-b bus Specifies the scsi bus for the device.
-d device number
Specifies the slot number for the drive to be removed or
inserted.
Additional options:
-t secs Specifies the number of seconds the command will wait while
the drive is inserted or removed. This default value is 300
(5 minutes). This is the amount of time the command will
wait before automatically continuing with the sequence. In
general the command continues when the user hits enter after
having inserted or removed the drive.
Page 2
scsiadminswap(1M)scsiadminswap(1M)-f Force the command to occur even if it discovers there are
still mounted file systems, or file systems in use. This is
very dangerous.
Examples
The following sequence of commands could be used to swap a disk in slot 2
with partition 7 mounted as /x
umount /x
scsiadminswap-u -b 0 -d 2
scsiadminswap-p -b 0 -d 2
mount -t xfs /dev/dsk/dks0d2s7 /x
The following sequence of commands could be used to swap a disk
containing old xlv information into slot 2 with partition 7 mounted as a
unmirrored xlv volume named /lv1.
umount /lv1
xlv_shutdown -n sample_vol
scsiadminswap-p -u -b 0 -d 2
xlv_mgr -xc "delete label /dev/rdsk/dks0d2vh" (delete old label information)
xlv_make (see xlv_make man page for examples)
xlv_mgr (perform attach)
mount -t xfs /dev/dsk/dks0d2s7 /lv1
SEE ALSOfuser(1M), mount(1M), umount(1M), xlv_assemble(1M), xlv_make(1M),
xlv_mgr(1M), xlv_shutdown(1M), fstab(4), mtab(4), scsiha(7M).
WARNINGS
Care should be take when using the scsiadminswap command. There are
periods when the scsi bus is in a sensitive state. The command should
always be allowed to continue to completion. Keep the amount of time it
takes to remove or insert the drive to a minimum. The most sensitive
time occurs during the unplug operation while the command is waiting for
the drive to be removed. During this period no other data can be written
to any device on the bus from which the drive is being removed. If the
drive can not be removed or inserted in the requested time do NOT kill
the command half way through, let it finish and re-try again later.
Killing the command during execution will leave the scsi bus in a bad
state. Also to help ensure command success and prevent the system from
failing, any file systems on the disk to be removed should not currently
be in use.
NOTE
The Performance Co-Pilot (see PCP(1)) may be used to monitor SCSI
devices, however configuration changes by scsiadminswap may not be
asynchronously detected. If SCSI devices are being monitored,
Page 3
scsiadminswap(1M)scsiadminswap(1M)libirixpmda(5) may be notified and forced to reset using the command:
> pmstore irix.xbow.switch 1
Page 4
