[Dirvish] [Bulk] Re: Full backup every time
Kenneth.Lerman at se-ltd.com
Fri Mar 21 23:25:55 UTC 2008
I believe that in the past, someone reported a similar problem. The cause
wound up being that he was doing a --init every backup.
That isn't your problem, is it?
Could you post the "summary" file from one of these backups?
----- Original Message -----
From: "Jens Lang" <jenslang at yahoo.de>
To: <dirvish at dirvish.org>
Sent: Friday, March 21, 2008 7:23 PM
Subject: Re: [Dirvish] [Bulk] Re: Full backup every time
Keith Lofstrom schrieb:
>>> what could be the reason that dirvish make a full backup every time I
>>> start the backup? I am running the lates Ubuntu on both, client and
> On Fri, Mar 21, 2008 at 05:24:46PM -0400, Kenneth Lerman wrote:
>> It is supposed to.
> Ken's reply is true, but you might want a little more information
> than that. Dirvish uses rsync to create full and complete images.
> On the server disk, there should be a succession of complete images.
> Rsync uses unix/linux hard links to share data file information; if
> a file does not change, there may be many links to it, but it is only
> stored on the backup disk once. Thus, a 500GB disk can store what
> appears to be hundreds of 200GB images - only the changed files and
> the directory information is added for each successive image. For
> example, I have a 500GB backup disk with about 150 images on it,
> each apparently 200GB, and I have used only 350GB of it. By using
> the "branch" feature of dirvish, files that are identical on many
> different machines can also share data space.
Yes, I know that it is supposed to do a full backup every time. Maybe my
wording was not exact. On my system it transfers the whole directory
every time (e. g. a large photo collection). These files never change.
And, of course, they use disk space on the backup disk.
> This is true in the ideal case. In reality, big files with little
> changes ( rotating-name logs, mbox mail folders, vmware images)
> will chew up disk pretty fast. Use "dateext" for logrotate,
> Maildir format for mail repositories, and samba links to linux
> filespace from smaller vmware images to minimize the big file
> changes getting backed up. Use dirvish-expire to manage storage.
> And look on the mailing list archives, the wiki, and the many
> FAQs and writeups created by other dirvish users for more hints.
> If after doing all these things, dirvish is creating huge images
> and not using hardlinks, filling up your backup disk too fast, there
> may be other problems. The helpful people on this list can help
> after you have made an effort characterize your problem.
The thing is that I have no idea how to find the cause. With another
computer it runs absolutely fine. I even don't know whether to search on
the client or on the server side.
The file system (for the backup) is ext3. I checked the size of the
directory with du–all backups have nearly 40 gb (except those who have
been interrupted). The remaining disk size (checked with df) decreases
by the same amount. I did not check the inode numbers, but I think the
decreasing disk size is proof enough.
On my client I am running a normal rsync server (version 2.6.9). The
server has enough rights to access the files.
I suppose that it for some reason thinks the file has changed although
it hasn't. But why? The file size and the date doesn't change.
Dirvish mailing list
Dirvish at dirvish.org
More information about the Dirvish