To create snapshot volumes of a filesystem, use the following procedure:

These procedures are described in “XVM Snapshot Administration”.

You can also perform the following procedures when administering XVM snapshot volumes:

The procedures are described in “XVM Snapshot Administration”.

This section provides information on the following tasks:

Setting up a Repository Volume

To set up a repository volume, first you create an XVM logical volume. A repository volume can have any legal XVM topography.

The size of the XVM volume that you will need will depend on several factors:

• The size of the filesystem for which you are creating a snapshot. A repository volume that is approximately 10% of this size could be a starting estimate.

• The volatility of the data in the volume. The more of the data that changes, the more room you will need in the repository volume.

• The length of time you will be keeping each snapshot before deleting it.

After you have created the XVM logical volume that you will use for the repository, you must initialize the repository volume. You can do this with the following command, where repository_volume is the name of the XVM logical volume:

xvm:local> repository -mkfs vol/repository_volume

---

Caution: Executing the -mkfs option destroys data. This command should be used only once, to create the repository filesystem.

---

This command issues an mkfs command of the repository volume. The repository command no longer mounts the repository. An open of any snapshot or base volume using that repository will cause it to get mounted. It gets closed again when the last snapshot or base using it gets closed.

xvm:local> vsnap -create repository vol/repository_volume vol/basevolume

vm:local> vsnap -delete repository vol/repository_volume vol/basevolume

xvm:local> show snapshot

xvm:local> show copy-on-write

df /dev/lxvm/volname

bayern2 # df /dev/lxvm/dks0d4s0
Filesystem             Type  blocks     use     avail  %use Mounted on
/dev/lxvm/dks0d4s0      xfs 35549600     1376 35548224   1  

xvm:local> repository -delete repository_vol

This section provides an example of basic snapshot configuration. It includes the following procedures:

# xvm
xvm:local> label -name disk0 dks2d70
disk0
xvm:local> label -name disk1 dks2d71
disk1
xvm:local> label -name disk2 dks2d72
disk2
xvm:local> slice -all disk*
</dev/lxvm/disk0s0>  slice/disk0s0
</dev/lxvm/disk1s0>  slice/disk1s0
</dev/lxvm/disk2s0>  slice/disk2s0
xvm:local> stripe -volname stripedvol slice/disk0s0 slice/disk1s0 slice/disk2s0
</dev/lxvm/stripedvol> stripe/stripe0
xvm:cluster> show -top stripedvol
vol/stripedvol                    0 online
    subvol/stripedvol/data     106627968 online
        stripe/stripe0             106627968 online,tempname
            slice/disk0s0              35542780 online
            slice/disk1s0              35542780 online
            slice/disk2s0              35542780 online
xvm:local> exit
# mkfs /dev/lxvm/stripedvol
meta-data=/dev/lxvm/stripedvol   isize=256    agcount=51, agsize=262144 blks
data     =                       bsize=4096   blocks=13328496, imaxpct=25
         =                       sunit=16     swidth=48 blks, unwritten=1
naming   =version 2              bsize=4096   mixed-case=Y
log      =internal log           bsize=4096   blocks=1632
realtime =none                   extsz=65536  blocks=0, rtextents=0

# mkdir /stripedvol
# mount /dev/lxvm/stripedvol /stripedvol

xvm:local> vsnap -create -repository reposvol stripedvol
repository name = reposvol
writing all SBs
new uuid = 9db5572c-5fe4-1028-8e9e-08006911cbcb

# mkdir /snap
# mount /dev/lxvm/stripedvol%0 /snap
# xfsdump -f dumpfile_path /snap

This section provides an example of a shell script that creates hourly snapshots of a filesystem. In this example, snapshots are retained for twenty-four hours, after which the oldest snapshot is deleted before a new snapshot volume is created.

Note that XVM snapshots do not provide backup against media failure; a full filesystem backup is necessary to protect against this. You can use snapshot filesystems in the event of data loss due to user errors such as accidental deletion.

---

Note: A snapshot volume name is the base volume name with %n appended, where n is the snapshot number, starting with 0. The value of n will continue to increase with each new snapshot. To begin the numbering scheme over again, you must delete all snapshots of a volume.

---

#!/bin/sh
#
#       Create hourly snapshot and mount it.
#
#       This script is intended to be called from a cronjob at certain times
#       per day.  It will cycle through 24 snapshots, allowing one snapshot per
#       hour.
#
#       Note that the name of the mount point is identified by the hour the
#       script is run.  If the script is run more than once in an hour, the
#       first snapshot created that hour will be unmounted and the new snapshot
#       will be mounted in its stead.
#
#       This script will take care of creating the mount directory and mount
#       points for the hourly snapshots, if they don't already exist.  It will
#       also not delete any snapshots until after the 24th snapshot.  Thus,
#       it's not necessary to run a different script for the first 24
#       snapshots.
#
#       To call this script:
#
#       sh ./snap basevol repvol regsize
#
#       where   basevol is the name of the base volume
#               repvol is the name of the repository volume
#               regsize is the name of the region size
#
#       Once the first snapshot has been created, the repvol and regsize
#       parameters will default to whatever was set for the first snapshot.
#
#       If regsize isn't specified for the first snapshot, it defaults to 128.
#
BASEVOL=$1
REPVOL=$2
REGSIZE=$3
MAXSNAPSHOTS=24
MOUNTDIR=/snapshot
XVM=/sbin/xvm
if [ “$XVM show -t vol/$BASEVOL 2> /dev/null | grep copy-on-write” ]
then
        COW=`$XVM show -t vol/$BASEVOL 2> /dev/null | grep copy-on-write | awk `{print $1}'`
        if [ “x$REGSIZE” = “x” ]
        then
                REGSIZE=`$XVM show -v $COW 2> /dev/null | grep “blks/reg:” | awk ` {print $10}'`
        fi
        FIRST=`$XVM show -v $COW 2> /dev/null | grep “first snap idx” | awk `{print $4}'`
        NEXT=`$XVM show -v $COW 2> /dev/null | grep “next snap idx” | awk `{print $8}'`
        if [ $FIRST -eq $NEXT ]
        then
                CUR=$NEXT
        else
                CUR=`expr $NEXT - 1`
        fi
        if [ $CUR -lt 0 ]
        then
                CUR=`expr $CUR + 65536`
        fi
else
        FIRST=0
        NEXT=0
        CUR=0
        if [ “x$REGSIZE” = “x” ]
        then
                REGSIZE=128
        fi
fi
SNAPSHOTS=`expr $NEXT - $FIRST`
if [ $SNAPSHOTS -lt 0 ]
then
        SNAPSHOTS=`expr $SNAPSHOTS + 65536`
fi

NEXTPATH=$1%$NEXT
CURPATH=$1%$CUR
FIRSTPATH=$1%$FIRST
OLDEST=/dev/lxvm/$FIRSTPATH
MOUNTPT=$MOUNTDIR/$BASEVOL.`date +%H`
NEWEST=/dev/lxvm/$NEXTPATH

# Now do the work.
#

# Make sure the mount point is there.
if [ ! -d $MOUNTDIR ]
then
        if [ -e $MOUNTDIR ]
        then
                echo “Snapshot directory $MOUNTDIR not a directory” >&2
                exit
        fi
        mkdir $MOUNTDIR
fi
if [ ! -d $MOUNTPT ]
then
        if [ -e $MOUNTPT ]
        then
                echo “Snapshot mount point $MOUNTPT not a directory” >&2
                exit
        fi
        mkdir $MOUNTPT
fi

# Unmount and delete the oldest snapshot, but only if we've already made the
# initial snapshots.
# Unmount $MOUNTPT in case we're out of sync
umount $MOUNTPT
if [ $SNAPSHOTS -ge $MAXSNAPSHOTS ]
then
        umount $OLDEST
        $XVM vsnap -delete vol/$BASEVOL
fi
# Now create a new snapshot.
$XVM vsnap -regsize $REGSIZE -repository vol/$REPVOL -create vol/$BASEVOL
mount $NEWEST $MOUNTPT

If your repository volume fills, you will not be able to perform I/O. You should take care to ensure that this does not occur. You can, however, determine in advance what action XVM should take in this circumstance, as described in “Determining System Behavior on Full Repository”.

You can use the df command to view the available repository space:

# df /dev/lxvm/reposvol
Filesystem             Type  blocks     use     avail  %use Mounted on
/dev/lxvm/reposvol      xfs 10690400      288 10690112   1  

If you find that your repository volume is filling up, you can grow your repository volume. To grow a repository volume, you add slices to the XVM logical volume by inserting a concat into the existing volume and then executing the xvm repository -grow command.

The following procedure grows the repository volume reposvol created in the procedure described in “Creating the Repository Volume”.

If your repository volume fills, you will not be able to perform I/O. This will likely lead to a filesystem shutdown when the xfs filesystem attempts to write metadata. You can, however, use the XVM change repfull command to determine in advance that XVM will delete the oldest snapshot of a volume when the repository volume fills, causing you to lose access to the snapshot but allowing the filesystem I/O to continue.

In short, when determining what the system should do when the repository volume fills, you can choose between one of these two options:

The following command configures the repfull parameter of the XVM volume stripedvol to specify that the oldest snapshot of stripedvol should be deleted automatically in order to free repository space when an I/O operation can't complete due to a full repository:

xvm:local> change repfull deloldest vol/stripedvol
vol/stripedvol

After setting this parameter, the oldest stripedvol snapshot will be deleted if the repository volume for stripedvol snapshots fills. Once enough space is freed, the I/O operation will continue. From that point, any I/O operation to the deleted snapshot will return an error, but the stripedvol filesystem will not shut down and there will be no xfs error message.

To specify that XVM return an error when I/O cannot be completed on stripedvol due to a full repository, use the following command:

xvm:local> change repfull error vol/stripedvol
vol/stripedvol

In this case, if the repository volume for stripedvol snapshots fills, XVM will return an error and it is likely that the stripedvol filesystem will eventually shutdown.

The default repfull parameter for an XVM volume is error. The repfull parameter is associated with an XVM volume, so you can set it even if you have not created any snapshots for that volume or if you have deleted all the snapshots for that volume.

When you create a snapshot volume with the XVM -vsnap -create command, the default size of the data regions that are copied to the snapshot volume is 128 blocks (64k). It may be possible to tune your system performance by setting the region size to a different value.

You can set the region size when you create the first snapshot of a volume with the -regsize n parameter of the vsnap -create command, where n is the desired region size in filesystem blocks (in units of 512 bytes).

You cannot change the region size after the first snapshot has been created. To change the region size, you must delete all snapshots of a volume (vsnap -delete -all vol) and create new snapshots.

You may want to take the following factors into consideration when determining the optimal region size for your snapshot:

You should take the following factors into consideration when you are using xfsdump to perform backup dumps of a snapshot filesystem:

You should take the following factors into consideration when you are using a third-party backup application package to perform backup dumps of a snapshot filesystem:

You should take the following factors into consideration when you are using either xfsdump or a third-party backup application package to perform backup dumps of a snapshot filesystem: