On 2015-08-03 16:54, Anton Aylward wrote:
On 08/03/2015 10:48 AM, Carlos E. R. wrote:
The name is the same, but the feature is distinctly different.
The purpose is to create a backup of the filesystem into a different device. It is not a COW (copy on write).
So it really generational backup for potential offlining, perhaps like disk-to-disk-to-tape.
From memory, and maybe faulty because I have never done it. I think the procedure is to block changes to the filesystem (this is the part I'm unsure about), then clone or image the filesystem somewhere else, as fast as possible, then free up changes again. If the filesystem would be able to freeze changes, but still accept them, written somewhere else, delayed, then applied later when the freeze is removed, that would be wonderful. I don't know if XFS is capable of this. Take a look at xfsdump and xfs_copy. xfsdump creates an image of an XFS filesystem, and does it very fast. It creates a single target file (an archive). It can be incremental. You do not need to umount it, I believe (I don't remember for sure; need more coffe). xfs_copy instead copies trees of files into another tree. Source and target must be XFS filesystems. Source should be be umounted, but could be "frozen" instead. It is similar to using rsync, but runs much faster, because it takes advantage of knowledge of the metadata structure. xfs_freeze - suspend access to an XFS filesystem The manual mentions that is intended for use with lvm or raid with snapshot support, but I don't see how. -- Cheers / Saludos, Carlos E. R. (from 13.1 x86_64 "Bottle" at Telcontar)