[Fedora-infrastructure-list] Backup thread

Curt Moore jcmoore at nuvio.com
Tue Jun 27 19:09:02 UTC 2006


On Tue, 2006-06-27 at 13:11 -0400, Jeffrey Tadlock wrote:
> Dennis Gilmore wrote:
> > Right now there are some cvs repositories, database dumps, and some other 
> > files  at about 34GB of data.   Id need someone with the right access to do a 
> > du -hs /home  on each server that has home dirs  needing backup.  to know how 
> > big that is.  
> 
> I spot checked a few of the boxes to see what their /home size was.  Of
> the app, proxy and db boxes had nothing over 36MB.  The internal box we
> use for our internal cvs repo had the largest /home at 728MB.  Probably
> a fair portion of that was just people's sandboxes.

As much as it pains me to mention it (as I think of NFS), would it be
worth looking at some sort of centralized storage so that the home
directories are the same across all boxes?  Likely NFS is not the best
solution here but it's the first that came to mind.

I had separate home directories on my systems a while back and we ran
into the same sort of debacle, not knowing on which box things were
stored, what had been backed up, managing profile settings across
machines, etc.  Since switching everything to use NFS it's made things a
lot easier from an admin standpoint.  Of course, I have a private GigE
LAN between all of the boxes so it makes this type of setup much easier.

If centralized home directories were coupled with a centralized account
management system, like with LDAP, it could be very powerful and ease
system administration.  This is, of course, the debate going on over in
the thread discussing the account system.

> 
> As I think more about it, rather than backup all /homes, how about we
> choose one that we let people know it is backed up and the others
> aren't.  As we work on scripts and such we know that if we expect it to
> be backed up that we keep it on the box that has /home backed up.  I
> would lean towards the box with the internal cvs repo that gets /home
> backed up (currently at 728MB).  What do people think about that?

Sounds like a good idea.  Would it be worth enforcing some user/group
disk quotas, if there aren't already, so that people don't knowingly or
inadvertently abuse the fact that it's somewhat of the central archive?
If so, at what values should these quotas set?

-Curt




More information about the infrastructure mailing list