veeamsnap kernel module required?

[gianluca]

Hello,
I would like to know if any kind of backup job type, such as file level backup to local repository (or to a share) implies presence of veeamsnap kernel module on the system.
In this case how intrusive is it? I don't like very much to have a layer in kernel to do basic (in my needed case) file level backup of some directories.
I'm trying the free edition and it seems that veeam is trying to create a volume snapshot anyway.
In case of needed kernel module, is there any clear tech spec about what it is doing in terms of system call and so on and leverage of which system functionalities?
Thanks
Gianluca

[PTide]

Hi,

All backup jobs require volume snapshot to be created and thus require veeamsnap module to be present. Without snapshot, file-level consistent cannot be guaranteed.

is there any clear tech spec about what it is doing in terms of system call and so on and leverage of which system functionalities?

Basically it relies on IO interception via substitution of standard bio processing function for block devices that are under change-block tracking with a custom one (tracking_make_request). Currently there is no tech.spec on veeamsnap module, only its source-code available via github.

Thanks

[gianluca]

Thanks for your reply.
But is the Veeam "snapshot" only at LVM level or also partition level if a file system is not on LVM?
Also, suppose that for data application insisting on a filesystem mounted on /a I already have a backup at application level, and I want to use Veeam agent for Linux to only backup the data on another file system mounted on /b.
If I configure the Agent to backup /b can I be sure that no I/O impact or snapshot related things would be present for /a? Or does it imply I/O impacts on all filesystems of the system as a whole?

[PTide]

But is the Veeam "snapshot" only at LVM level or also partition level if a file system is not on LVM?

Veeam Agent does not rely on LVM snapshots, yet it has LVM2 packages in dependencies for compatibility with LVM partitions. Instead of LVM snapshot Veeam uses its own snapshot mechanism implemented in veeamsnap kernel module.

Also, suppose that for data application insisting on a filesystem mounted on /a I already have a backup at application level, and I want to use Veeam agent for Linux to only backup the data on another file system mounted on /b

You can select any partition (except for those that belong to RAID or VG) or mount point for backup (given that the filesystem is supported by Veeam Agent), no need to select the whole machine for backup.

If I configure the Agent to backup /b can I be sure that no I/O impact or snapshot related things would be present for /a?

Veeam chooses a place to store snapshot data automatically, however you can specify that location manually in /etc/veeam/veeam.ini. That will ensure that all COW I/O won't go to /a.

Thanks