[SGVLUG] backups -- was Dev-sig location

Claude Felizardo cafelizardo at gmail.com
Fri Apr 24 12:48:03 PDT 2009


On Thu, Apr 23, 2009 at 12:02 PM, David Lawyer <dave at lafn.org> wrote:
> On Thu, Apr 23, 2009 at 04:51:11PM -0700, Eric Hammond wrote:
>>
>> For backups, I use dirvish which is also based on rsync and hardlinks.
>> Claude Felizardo wrote:
>> > I use rsnapshot which is based on Mike Rubel's original set of
>> > scripts.
>
> I haven't yet found a satisfactory backup.  I'm now using the old and
> unmaintained cpbk (stands for copy backup).  What it does is syncs my
> 2nd hard disk to look like the 1st one, at least for /home and certain
> other important stuff.  Since the 2nd hard disk is bootable, should
> the 1st disk fail, I would just boot from the 2nd disk and have all my
> files the way they were the previous day.  It also has a trash
> directory on the 2nd hd for saving deleted files.
>
> But a problem is: suppose I accidentally delete part of a file (on the
> 1st disk) but don't realize it.  Then I keep modifying the file and
> the original one in the trash gets overwritten so I loose the stuff I
> accidentaly deleted.  I would like the trash directory to append
> numerals to file name such that putting a file in the trash directory
> never deletes the older files with the same base name.
>
> Of course I could use both a sync program like unison to keep the hard
> disks in sync and then use a backup program like rsnapshot to protect
> against accidental corruption of files.  But it would be easier if one
> program did it all (like cpbk tries to do).
>
>                        David Lawyer
>

What I like about rsnapshot is that it saves multiple versions so I
can go back and pick which version I want but if the file has not
changed, it uses a hard link so the only additional space used is an
inode.

As a developer I'm constantly making changes to source code (or even
system config files on the machines that I maintain) and I find it
extremely helpful to be able to go back and grab a version from hours
or days or weeks ago w/o having to go to the sysadmins and ask for a
restore.  Of course we use a code repository for stuff that has been
tested but if I'm working on something and get distracted for a few
days and then decide that what I was working on wasn't going to work
and I want to look at what I had before going on vacation, I can do
that with rsnapshot.

At work, I run 12 hourly backups from 10 am to 8 pm plus midnight and
a daily backup going back 40 days.  At home, I only have 6 hourly
backups running every 4 hours but the dailies go back 90 days.
Rsnapshot will automatically delete the oldest set for each given
interval.  So as long as I don't make lots of big changes the total
disk usage doesn't change too much over time.

Disclaimer, rsnapshot is not a substitute for regular backups.

What I'd really like is a versioning filesystem.  Dec had something
that did that, Was it VMS?  Each version was accessible with a ';n'
appended to the filename which was automatically incremented so for
example test.txt;1 was the previous version, test.txt;2 was the next
and test.txt was the same as test.txt;0.  Not sure as it's been
decades since I used one.

claude


More information about the SGVLUG mailing list