pclissold wrote:
You'll get a copy where open files are in undefined states - as if the machine had been suddenly powered off. If you are using a journalled filesystem, this shouldn't matter in theory, but the copy is not instantaneous like a power-down so the results are hard to predict. Since this will be the backup you rely on after the original has been trashed, play it safe and shut down while you duplicate.
No, this is not the same as a sudden poweroff. Since the linode is modifying the filesystem at the same time, you can (and probably will) end up with massive corruption - and your journal will only make things worse, as the kernel might not notice the corruption until it's already made it worse. If for some reason such a thing was /absolutely necessary/ for some insane reason:
* Quiesce the source filesystem as much as possible first
* Copy the fs
* Forcibly remove the journal from the copy with debugfs
* Run e2fsck -f
* Pray (but don't get your hopes up)
In order for the copy to truly be sound, it absolutely needs to be made atomically.