Hi Adam,
Thanks for the quick response but I doubt I can split 16TB of data and then
backup in a timely manner.
But it's worth a try.
Cheers,
Pedro
On Friday 04 March 2011 13:27:13 Adam Goryachev wrote:
> On 04/03/11 23:32, Pedro M. S. Oliveira wrote:
> > I'm not sure about doing the 16TB (performance, backup duration) so I
> > thinking in some kind of block device backup.
> > Idea:
> > 1 - Create lvm snapshot of the block device
> > 2 - Backup lvm snapshot (I could use DD, but then it would be a full backup
> > every time I do a backup), something like rsync where the only the changed
> > blocks of the block device.
> >
> > Benefits:
> > 1 - Performance, althoughtthe gains only show after 70% of full disk,
> > 45%-50% full disk for small files.
> > 2 - Restore backup directly into volume.
> > 3 - Possibility of mount on a loop device.
> >
> > Conns:
> > The first backup should take ages, and initial FS should have zeros on it's
> > free space (so the initial backup can use the compression efficiently)
> > This approach is only possible on unix/linux FS.
> > The LVM methods for creating snapshots aren't standard and partitioning /
> > volume creation need to be addressed and thought before deployment (is this
> > a conn??)
> >
> > The recover method should be able to restore the block device (in this case
> > an LVM volume).
> > I can see lots of difficulties with this approach but the benefits can be
> > great too.
> > What do you all think about this.
>
> Been there, done that, and it works well already (sort of)...
>
> I have a MS Windows VM which runs under Linux and it's 'hard drive' is
> in effect a file. I used to:
> 1) use LVM to take a snapshot
> 2) copy the raw snapshot image to another location on the same disk
> 3) delete the snapshot
> 4) split the copy of the image to individual 20M chunks (split)
> 5) use backuppc/rsync to backup the chunks
>
> The problem with this is the time for the backup to complete (about
> hours for steps 1 - 3, and another 1 hour for step 4
>
> Recently, I skipped step 1 and 3 and just shut down the machine before
> taking the copy, this finishes in about 30 minutes now. In any case,
> backuppc handles this quite well.
>
> The reasons for splitting the image are:
> 1) backuppc can take better advantage of pooling since most chunks have
> not changed (one big file means the entire file changes every backup).
> 2) backuppc doesn't seem to handle really large files with changes in
> them (ie, performance wise it seems to slow things down a lot).
>
> Hope this helps...
>
> PS, I also use the same idea for certain niche applications that produce
> a very large 'backup' file, I split it into chunks before letting
> backuppc back it up. I also ensure they are de-compressed first, letting
> rsync/ssh/backuppc handle the compression at the transport and file
> system levels.
>
> Regards,
> Adam
>
> --
> Adam Goryachev
> Website Managers
> www.websitemanagers.com.au
>
> ------------------------------------------------------------------------------
> What You Don't Know About Data Connectivity CAN Hurt You
> This paper provides an overview of data connectivity, details
> its effect on application quality, and explores various alternative
> solutions. http://p.sf.net/sfu/progress-d2d
> _______________________________________________
> BackupPC-users mailing list
> BackupPC-users AT lists.sourceforge DOT net
> List: https://lists.sourceforge.net/lists/listinfo/backuppc-users
> Wiki: http://backuppc.wiki.sourceforge.net
> Project: http://backuppc.sourceforge.net/
>
--
----------------------------------------------------------------------------------------------------------
Pedro M. S. Oliveira Pólo Tecnológico de Lisboa
IT Consultant Estr. do Paço do Lumiar,
Lote 06, Edifício Multitech - 2C1
Email: pedro.oliveira AT dri DOT pt 1600-546 Lisboa
URL: http://www.dri.pt http://www.linux-geex.com
Telefone: +351 21 715 30 55 Fax: +351 21 715 30 57
----------------------------------------------------------------------------------------------------------
------------------------------------------------------------------------------
What You Don't Know About Data Connectivity CAN Hurt You
This paper provides an overview of data connectivity, details
its effect on application quality, and explores various alternative
solutions. http://p.sf.net/sfu/progress-d2d
_______________________________________________
BackupPC-users mailing list
BackupPC-users AT lists.sourceforge DOT net
List: https://lists.sourceforge.net/lists/listinfo/backuppc-users
Wiki: http://backuppc.wiki.sourceforge.net
Project: http://backuppc.sourceforge.net/
|