The method of customizing the shell script for backup pre- and post-processing is explained below.
When you do not want to unmount a transaction volume
Edit the script to delete the comment characters ("#") on lines 90-98, 102 and 135 for a transaction volume that is mounted and that you do not want to unmount. Then specify the target device in the if statement on line 90.
This operation is possible only when the file system is one of the following two:
ufs
SynfinityFILE (single-partition configuration)
The pre-processing by the customized scripts is as follows:
State of transaction volume | preprocessing |
|---|---|
It is mounted. | Updating to operating volume is deterred. |
The post-processing by the customized scripts is as follows:
preprocessing | post-processing |
|---|---|
If Update to operating volume was prevented. | Restore capacity to update operating volume. |
Note
Where the standard script is used, the caching of a file system and renewal of subsequent updates are inhibited by unmounting a file system before an advanced copy process using (EC/OPC), the integrity of the data of the entire file system is guaranteed.
Although flushing the cache of a file system and renewal of subsequent updates are inhibited using "lockfs -w", when customizing a script to prevent unmounting a file system, data integrity or file consistency cannot be guaranteed if the file is in Write Open mode (the database space of Oracle which is supporting online backup).
Moreover, when the renewal of metadata is occurring, the data updates to the file system itself are not guaranteed even with "lockfs -w" specified.
It is therefore critical to consider this and to manage the integrity of data updates accordingly when you elect not to unmount a file system.
If a SDX object is operated in slice units, do not specify slice name (/dev/sfdsk/<class name>/dsk/<volume name>:<sdxinfo DEVNAM value>) in Line 90. Instead, specify logical volume name (/dev/sfdsk/<class name>/dsk/<volume name>).
Cluster operation by VERITAS Cluster Server
Customization of the pre-processing and post-processing scripts is required when performing cluster operations using VERITAS Cluster Server, and when the mount point of the transaction volume is registered in the cluster transaction.
Change the unmount/mount processing in the pre-processing and post-processing scripts into offline/online processing of the mount point resource.
The mount point resource relating to the Volume is immediately offline/online whilst there will be a time delay before access is actually prevented when mount/unmount is used.
Please note that once the online/offline command is issued there will be a delay before the mount/unmount command can be executed. The sleep and df commands can be used to delay the mount/ unmount until after the process that sets the volumes online or offline has completed.
The examples below show how the pre-processing and post-processing scripts are modified.
[lines 104 and 113]
(before change) | /usr/sbin/umount $mount_point |
(after change) | /opt/VRTSvcs/bin/hares -offline resource-name -sys system-name |
[line 133]
(addition) | while /usr/sbin/df -l "$device">/dev/null 2>&1; do :; done |
[lines 112, 116, 118, 131, 135 and 137]
(before change) | /usr/sbin/mount $device $mount_point or /usr/sbin/mount -F $fs $device $mount_point |
(after change) | /opt/VRTSvcs/bin/hares -online resource-name -sys system-name |
[line 157]
(addition) | until /usr/sbin/df -l "$device">/dev/null 2>&1; do :; done |
