DOC HOME SITE MAP MAN PAGES GNU INFO SEARCH PRINT BOOK
 

vxplex(1M)


vxplex - perform Volume Manager operations on plexes

Synopsis

vxplex [ -Vf ] [ -g diskgroup ] [ -U usetype ] [ -o useopt ] att volume plex ...
vxplex [ -Vf ] [ -g diskgroup ] [ -U usetype ] [ -o useopt ] [ -v volume ] det plex ...
vxplex [ -Vf ] [ -g diskgroup ] [ -U usetype ] [ -o useopt ] [ -v volume ] dis plex ...
vxplex [ -Vf ] [ -g diskgroup ] [ -U usetype ] [ -o useopt ] cp volume plex ...
vxplex [ -Vf ] [ -g diskgroup ] [ -U usetype ] [ -o useopt ] snapstart volume plex
vxplex [ -Vf ] [ -g diskgroup ] [ -U usetype ] [ -o useopt ] snapshot plex newvolume
vxplex [ -Vf ] [ -g diskgroup ] [ -U usetype ] [ -o useopt ] [ -v volume ] mv oldplex newplex

Description

The vxplex utility performs Volume Manager operations on plexes and on volume-and-plex combinations. The first operand is a keyword that determines the specific operation to perform. The remaining operands specify the configuration objects to which the operation is to be applied.

Each operation can be applied to only one disk group at a time, due to internal implementation constraints. Any volume or plex operands will be used to determine a default disk group, according to the standard disk group selection rules described in vxintro(1M). A specific disk group can be forced with -g diskgroup.

These are the recognized operation keywords:

att
Attach each named plex to the named volume. This can be applied to dissociated plexes, or to non- enabled plexes already associated with the named volume. If the volume is enabled, then the result of the successful operation will be to associate the plex (if needed) and to recover the plex to have the same contents as all other attached plexes in the volume. The rules for performing the attach depend upon the usage type of the named volume.

Attaching a plex is the normal means of recovering a plex after a disk replacement, or after a plex offline.

det
Detach each of the named plexes. Detaching a plex leaves the plex associated with its volume, but prevents normal volume I/O from being directed to the plex. This operation can be applied to plexes that are enabled or disabled. The rules for performing the detach depend upon the usage types of the volumes involved. The operation does not apply to dissociated plexes.

dis
Dissociate each of the named plexes. Dissociating a plex breaks the link between the plex and its volume. A dissociated plex is inaccessible until it is reassociated, which can be done either with vxplex att or with vxmake. Any checks and synchronizations that apply to the det operation also apply to the dis operation.

Plex dissociation is the normal means of unmirroring a volume, or reducing the mirror count for a volume. To support this use, -o rm can be used to dissociate and remove the plex (and its associated subdisks) in the same operation. This makes the space used by those subdisks usable for new allocations (such as with vxassist or with vxmake).

Plex dissociation can also be used for file system backups of volumes that are normally mirrored. Plex devices are not directly mountable, so the backup method described for the det operation will not work if the backup program requires a mounted file system. To support such backup programs, a plex can be dissociated and can then be allocated to a new volume, such as with the command:

	vxmake -U gen vol volume plex=plex 
The created volume can then be started and mounted for use by the backup program.

cp
Copy the named volume to the named plexes. The volume cannot be enabled, and the named plexes must not be associated. The results of the operation will be a set of dissociated plexes that are an exact copy of the volume at the time of completion of the operation. The rules for performing the attach depend upon the usage type of the named volume. To improve the quality of the copies, some usage types attempt to make the detached plex consistent with respect to in-memory data.

This operation can be used to make a copy of a volume, for backup purposes, without mirroring the volume in advance.

snapstart and snapshot
These two operations form the two parts of a preferred means of copying a volume to a plex for backup purposes. The snapstart operation attaches a plex to a volume and, when the operation is complete, leaves the plex associated as a temporary plex. After the operation completes, the administrator can convert the plex attached by snapstart into a new volume using vxplex snapshot. To improve the quality of the copies, some usage types attempt to make the detached plex consistent with respect to in-memory data.

This method of backup is preferable to using vxplex cp because it allows the administrator to coordinate breaking off the plex from the original volume at a well-defined point in time. This is important, since attaching a plex to a volume can take a considerable amount of time, and it is difficult to know when it will complete. Also, directly converting the plex into a new volume is more convenient than requiring additional steps.

mv
Attach the plex newplex to the volume that oldplex is associated with and dissociate oldplex. The volume cannot be disabled, and newplex must name a dissociated plex. The operation ensures seamless replacement of the dissociated plex without loss of data in the volume and without significant delays in volume accessibility.

A primary purpose for the plex move operation is to move a plex that is using a disk to another location. In support of this purpose for the operation, -o rm can be specified to remove the original plex after completion of the operation.

For concatenated or striped plexes, the vxsd mv operation can be used to move individual subdisks off a disk. The rules for performing the move depend upon the usage types of the volume to which oldplex is associated.

Options

-g diskgroup
Specify the disk group for the operation, either by disk group ID or by disk group name. By default, the disk group is chosen based on the name operands.

-U usetype
Limit the operation to apply to this usage type. Attempts to affect volumes with a different usage type will fail.

-o useopt
Pass in usage-type-specific options to the operation. A certain set of operations are expected to be implemented by all usage types:

slow[=iodelay]
Reduce the system performance impact of copy operations. Copy operations are usually a set of short copy operations on small regions of the volume (normally from 16 kilobytes to 128 kilobytes). This option inserts a delay between the recovery of each such region. A specific delay can be specified with iodelay as a number of milliseconds; otherwise, a default is chosen (normally 250 milliseconds).

iosize=size
Perform copy operations in regions with the length specified by size, which is a standard Volume Manager length number (see vxintro(1M)). Specifying a larger number typically causes the operation to complete sooner, but with greater impact on other processes using the volume. The default I/O size is typically 32 kilobytes.

rm
Remove the plexes after successful completion of a vxplex dis operation. Remove the source plex after successful completion of vxplex mv.

-v volume
Require that the plex named by a plex or oldplex operand be associated with the named volume. This option can be used as a sanity check, to ensure that the specified plex is actually the plex desired for the operation.

-V
Display a list of utilities that would be called from vxplex, along with the arguments that would be passed. The -V option performs a ``mock run'' so the utilities are not actually called.

-f
Force an operation that the Volume Manager considers potentially dangerous or of questionable use. This permits a limited set of operations that would otherwise be disallowed. Some operations may be disallowed even with this flag.

Fsgen and gen usage types

The fsgen and gen usage types provide similar, though not identical, semantics for all operations of the vxplex utility. In particular, the fsgen usage type will attempt to flush in-memory data cached for the file system residing on the volume. For most file systems, this consists of calling sync(1M) to attempt to flush all in-memory data to disk. For the vxfs file system type, this will use special ioctls to ensure a reliable flush of the involved volume.

If a vxplex operation is interrupted by a signal, then an attempt is made to restore the disk group configuration to a state that is roughly equivalent to its original state. If this attempt is interrupted (such as through another signal) then the user may need to perform some cleanup. The specific cleanup actions that are needed are written to the standard error before vxplex exits.

The fsgen and gen usage types provide the following options as arguments to -o in addition to the required options:

force
Force an operation that the Volume Manager considers potentially dangerous or of questionable use. This applies to attempts to detach or dissociate the last (complete) plex in a volume, or to attempts to move a plex to a plex that has a different size. This flag is the same as -f.

rerr
Ignore volume or plex read errors when copying data onto a plex. A warning message is written to standard error if a read error occurs, but the error does not affect success of the operation. This operation can be used only with the cp operation; the operation is ignored if used with other operations.

werr
Ignore plex write errors when copying data onto a plex. A warning message is written to standard error if a write error occurs, but the error does not affect success of the operation. This operation can be used only with the cp operation; the operation is ignored if used with other operations.

mapzero
If a plex is moved to a new plex that has regions that are mapped to a subdisk in the destination, but are not mapped to a subdisk for any enabled, readable plex in the volume, then zero out that mapped region in the destination plex. Without this flag, the mapped region may be left unchanged from its original contents.

Limitations and extensions for the fsgen and gen usage types consist of the following:

att
If the volume is enabled and one of the named plexes is associated with the volume, then the plex must be STALE, EMPTY, ACTIVE, or OFFLINE. If the operation succeeds in attaching a plex, then any I/O fail condition for the plex is cleared. Also, attaching to an enabled volume requires that the volume have at least one enabled, read-write plex.

If the volume is not enabled, then the named plexes are associated with the volume (if not already associated) and are set to the STALE state, so that the plex will be fully attached by the next vxvol start or vxvol startall operation that applies to the volume.

If the log type of the volume is UNDEF and an unassociated plex with a log subdisk is attached, the volume is automatically converted to have a log type of DRL. Logging of volume changes is enabled when the volume has at least one enabled, associated plex with an enabled log subdisk and at least two read-write mode plexes.

An attempt to attach an unassociated plex fails if the putil0 field is not empty. This makes it possible to prevent use of a plex by using vxedit set to set the putil0 field to a non-empty string. The putil0 field can then be cleared with either vxedit set or with vxmend clear putil0.

dis and det
A detach or dissociate of a plex in an enabled volume fails if applied to a plex that is the last complete, enabled, read-write plex in the volume and the volume contains two or more non- complete, enabled, read-write plexes. In other words, a volume cannot be left with two enabled, non-complete plexes. A complete plex is one that is at least as long as the volume, and has subdisks mapped to the plex for all blocks up to the length of the volume. The -f option is required to reduce a volume to containing one enabled, read-write, non-complete plex, or to having no enabled, read-write plexes at all.

The det operation changes the state for an ACTIVE or CLEAN plex to STALE. The next time the volume is started, the plex will be re-attached automatically.

cp
The fsgen and gen usage types do not add any specific restrictions to the cp operation.

mv
If the destination plex has unmapped regions (a range of blocks in the plex with no backing subdisk) that are not mapped in the source plex, or if the destination plex is shorter than the source plex, then the -f option is required. Even with -f, the operation will prevent the plex from being sparsed such that the volume would be left with two or more sparse, enabled, read- write plexes, but no complete plexes.

RAID5 usage-type

The raid5 usage type provides the following options as arguments to -o in addition to the required options:

force
Force an operation that the Volume Manager considers potentially dangerous or of questionable use. This applies to attempts to dissociate the RAID-5 plex of a non-EMPTY volume or to remove the last RAID-5 log plex of a non-EMPTY volume.

As with other usage types, if a vxplex operation is interrupted by a signal, then an attempt is made to restore the disk group configuration to a state that is roughly equivalent to its original state. If this attempt is interrupted (such as through another signal) then the user may need to perform some cleanup. The specific cleanup actions that are needed are written to the standard error before vxplex exits.

The raid5 usage type supports only the following keywords:

att
Attach the named plexes to the named volume. If a plex has a layout of RAID, the plex will be attached as the RAID-5 plex of the RAID-5 volume. To attach a RAID-5 plex to the volume, the volume must be disabled and be in the EMPTY state, and the RAID-5 plex will be given a state of EMPTY.

If a plex has a layout other than RAID, the plex will be attached as a RAID-5 log plex for the RAID- 5 volume. If the volume has no RAID-5 log plexes, the log length for the volume will be set to the length of the smallest log plex being attached. If the volume already has at least one log plex, a plex can only be attached as a log plex if its contiguous length is at minimum the volume's log length. RAID-5 log plexes cannot be sparse in respect to the volume's log length; attempts to attach a sparse log plex will fail.

If the RAID-5 volume is not enabled, log plexes are attached and marked as STALE. If the RAID- 5 volume is enabled and has no log plexes, attaching a log plex will cause plexes being attached as log plexes to be zeroed before they are enabled. Otherwise, the new log plexes are attached write-only and the contents of the existing log plexes are copied to the new log plexes using ATOMIC_COPY ioctls, after which the logs are enabled.

dis
Dissociates the named plex from the RAID-5 volume to which it is attached. If the plex is the RAID- 5 plex of the volume and the volume is not EMPTY, this requires the -o force option, as any data on the volume would be lost. If the plex is a log plex for the volume and will leave the RAID-5 volume with no usable log plexes, the -o force option is required.

Note that the RAID-5 usage type does not support the det, cp, snapstart, snapshot or copy keywords; these operations are either inappropriate or impossible to perform within the operational concepts of RAID-5.

Files

/etc/vx/type/usetype/vxplex
The utility that performs vxplex operations for a particular volume usage type.

/etc/vx/type/fsgen/fs/fstype/vxsync
Path to a program used with the fsgen usage type for synchronizing in-memory file system data with a volume, for the file system type fstype. The program is given arguments of a volume name and one or more plex names. For the ufs and s5 file system types, this is a link to sync. For the vxfs file system type, this program uses the VxFS file system freeze feature to ensure a perfect synchronized detach.

Exit codes

The vxplex utility exits with a nonzero status if the attempted operation fails. A nonzero exit code is not a complete indicator of the problems encountered but rather denotes the first condition that prevented further execution of the utility. See vxintro(1M) for a list of standard exit codes.

References

sync(1M), vxassist(1M), vxedit(1M), vxintro(1M), vxmake(1M), vxmend(1M), vxsd(1M), vxvol(1M)



© 1997 The Santa Cruz Operation, Inc. All rights reserved.