automatic hard linking

[Tom Buskey]

On Wed, Jul 23, 2008 at 1:48 PM, <t.littlefield at comcast.net> wrote:

> Hello,
>
> A co-worker and I were talking about various ways to do 'backups' to try
> and
> prevent data loss.  The topic came around to a file system we had used
> at a previous job.  I can't remember the specifics, but we believe it was
> a Network Appliance system.
>
> One of the cool features it offered was a series of hourly, nightly and
> a monthly backup of files.  We kind of surmised that it was some sort of
> hard linking of the same file name in a different directory...  i.e.
>
> ~/foo.txt
>
> hourly.0/~/foo.txt
>
> So, if you accidentally deleted your home directory copy, you could still
> recover
> it for a short time.  Once it made it past X hours, the file would start
> getting
> removed from the hourly.n directories.  We were never admins on the box
> nor do we have experience with it, so the hard link is just a theory...



The snapshotting is part of the WAFL file system and the NetApp OS: OnTap.

ZFS has a similar snapshotting.  WAFL and ZFS are both COW filesystems and
that makes this kind of thing easier.

ZFS doesn't have the automatic snapshotting feature built in.  However,
someone has created a Solaris SMF thingy that will essentially cron it.  You
pick frequency and how many copies you want to keep around.

Other file systems do snapshotting, but they're not setup to keep more then
1 copy around AFAIK.  Solaris UFS can do a snapshot and I think LM can (I'm
fuzzy on this).  The idea is to stop your app, snapshot, start your app,
backup the snapshot, blow the snapshot away.  It's great for database
backups.



> A quick google search on automatic hard linking didn't turn anything up.
>  Is
> there a package anyone knows of that registers for file system events and
> will create temporary hard links in another location for 'accident
> prevention'.
>
> This isn't a backup strategy (obviously) just a method for easily
> recovering
> files when that "Oops" happens...


I agree with you there.  I'd bet netapp style snapshotting will cover most
of a typical sysadin restore needs.

I had 3 NetApps for >100 users over a 4 year time.  I had disk and
motherboard failures but never had to restore from tape in all that time.
RAID4 covered the disk failures, snapshotting the user "Oops" and blind luck
covered the motherboard failure.

I think it's important to think about data protection as multiple parts:

1) Oops!
   Snapshots will help here

2) Disaster!  The hardware/site failed; Get me the most current data set
    Backup to another media for this.  Tape, a 2nd file server
    These get recycled and over written

3) Archiving!  I want a file from 6Jan1999
     DVDs, HSM, Centerra or Bitlocker (?) servers, a tape before the data is
deleted.
     Your backup server *can* do this, but it's really a different focus.
     Many people confuse archiving with backups.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.gnhlug.org/mailman/private/gnhlug-discuss/attachments/20080723/98625303/attachment.html