On Fri, Mar 10, 2006 at 01:39:24PM -0600, Les Mikesell wrote:

> Rsync incrementals turn on the option to skip files where the length
> and timestamp match, making them considerably faster but more likely
> to accidentally miss a changed file. However, since they always work
> against the last full, you end up re-copying things that were copied
> in previous incrementals.  There is some room for improvement here.  

A while back, I tried using Retrospect backup's linux client for backup.
Aside from quickly discovering that they weren't prepared to deal with some
of the long pathnames and filenames on my machine, I discovered a much more
fundamental problem.

By default, retrospect tries to do something similar to backuppc, in that
files from previous backups aren't copied again, but just refer to the old
file.  Essentially, it is kind of like backuppc, with rsync timestamps and
always doing full backups (retrospect doesn't expire old data, you have to
start a new 'full' backup to do that).

However, I discovered they were just using file timestamp and size to
indicate duplicates.  That meant that any file of the same size and
timestamp would be considered the same (no hash).  It completely mangled
the gentoo portage tree, since it gets rsync updated, and has many files
that just consist of a few MD5 hashes.  Large groups of these files are the
same size, and are updated within a second.  Very bad assumption for them
to make.

What is much better is to only consder the timestamp for the same given
file (like rsync is doing).  This at least prevents merging arbitrary
files.

Also, as far as I know, every backup program that does incrementals looks
at just the timestamp field.  On unix, the ctime and mtime field together
will always tell you of a file change, as long as the system clock is
monotonically increasing.  On windows, all bets are off, and I'm not sure
there really is anything to do but read the file in and hash it's entire
contents.

Dave


-------------------------------------------------------
This SF.Net email is sponsored by xPML, a groundbreaking scripting language
that extends applications into web and mobile media. Attend the live webcast
and join the prime developer group breaking into this new coding territory!
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=110944&bid=241720&dat=121642
_______________________________________________
BackupPC-users mailing list
BackupPC-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/backuppc-users
http://backuppc.sourceforge.net/

Reply via email to