On 5/17/07, Mark Houts <mhouts@highwire.stanford.edu> wrote:
> I'm backing up an FSFS repository that is large (100GB+) and heavily
> used by geographically dispersed groups so there's no obvious time to
> schedule a backup when the repository is not being accessed. I'd like to
> do hot copies while the repository is in use but I am concerned about
> whether any transactions taking place during the backup could be
> corrupted. Is there any functional difference between hotcopy and rsync
> either with the resulting backup repository or how open files are handled?
We have a large FSFS repository too -- it was about 100GB, but it went
down to 40GB when we upgraded the server to svn 1.4.3. I recommend
multiple points of attack for backup, in general. So I hotcopy FSFS to
such a drive, mounted locally on the same system serving svn. I
immediately follow that with an "svn dump", and then I rsync the
hotcopy-ied copy to a remote server. This gives us redundant local
backup, and off-site remote backup. Hotcopy is safer, and you can
always rsync the copy later. Having a dump around is good for server
hardware changes and software upgrades, which can change the FSFS
format, but its output much slower and bigger.
500GB+ SATA drives are currently cheap backup devices, which can
be mounted via USB, Firewire, eSATA, or hot-swap tray to internal
SATA, configured for hot removal. If configuring a SATA drive for
removal is too much trouble, mount it permanantly, and use it as a
way-point between the original data, and some other removable backup
media. In my opinion, there's currently no good excuse for not keeping
a local backup, and backup over a network connection (like rsync)
should only be used for redundant off-site network backup.
:) Jred
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Received on Thu May 17 22:30:36 2007