On Mon, 10 Sep 2001, Takayuki Sasaki wrote:
> I would like to know that whether it is correct way or
> workaround to dump the XFS file system which requires preceding
> unmount and mount, not only for QA suite but also on a
> production environment.
> Could you explain me about this?
For a production environment, it would not be necessary to do the
umount and mount to stabilise the filesystem.
The fact is that xfsdump must be run on a mounted filesystem, which
means there is always the possibility that it will miss files which
have been created immediately before the execution of xfsdump and
during the execution of xfsdump. These files will be caught in the
backup the next time you run xfsdump. There is nothing xfsdump can do
to protect against this situation -- it is up to the system
administrator to try to run xfsdump when the system is not busy. If
the system is always very busy, then the administrator should be sure
to run xfsdump on a regular basis.
In the QA environment, we create a test filesystem and immediately dump
it, so we need to take special steps to ensure the test works as
Basically, it's simply a matter of understanding that the backup
created by xfsdump is a snapshot of the filesystem as it was 30 seconds
ago, rather than the moment xfsdump was run. For most people, I
would expect this would be fine.
 30 seconds is just an estimate. This is reliant on the new inode
data being being flushed to disk ... I'm not sure how long it takes
normally. Someone else on the list might have a better idea.