Veritas Storage Foundation 5.1 SP1 Advanced Features Administrator"s Guide (5900-1503, April 2011)

level of the directory. The vxfilesnap command preserves the inode identify of
the destination file, if the destination file exists.
All regular file operations are supported on the FileSnap, and VxFS does not
distinguish the FileSnap in any way.
See the vxfilesnap(1) manual page.
Properties of FileSnaps
FileSnaps provide an ability to snapshot objects that are smaller in granularity
than a file system or a volume. The ability to snapshot parts of a file system name
space is required for application-based or user-based management of data stored
in a file system. This is useful when a file system is shared by a set of users or
applications or the data is classified into different levels of importance in the
same file system.
FileSnaps provide non-root users the ability to snapshot data that they own,
without requiring administrator privileges. This enables users and applications
to version, backup, and restore their data by scheduling snapshots at appropriate
points of their application cycle. Restoring from a FileSnap is as simple as
specifying a snapshot as the source file and the original file as the destination file
as the arguments for the vxfilesnap command.
FileSnap creation locks the source file as read-only and locks the destination file
exclusively for the duration of the operation, thus creating the snapshots
atomically. The rest of the files in the file system can be accessed with no I/O
pause while FileSnap creation is in progress. Read access to the source file is also
uninterrupted while the snapshot creation is in progress. This allows for true
sharing of a file system by multiple users and applications in a non-intrusive
fashion.
The name space relationship between source file and destination file is defined
by the user-issued vxfilesnap command by specifying the destination file path.
Veritas File System (VxFS) neither differentiates between the source file and the
destination file, nor does it maintain any internal relationships between these
two files. Once the snapshot is completed, the only shared property between the
source file and destination file are the data blocks and block map shared by them.
The number of FileSnaps of a file is practically unlimited. The technical limit is
the maximum number of files supported by the VxFS file system, which is one
billion files per file set. When thousands of FileSnaps are created from the same
file and each of these snapshot files is simultaneously read and written to by
thousands of threads, FileSnaps scale very well due to the design that results in
no contention of the shared blocks when unsharing happens due to an overwrite.
The performance seen for the case of unsharing shared blocks due to an overwrite
Administering FileSnaps
Properties of FileSnaps
294