Re: Re: Incremental backup including entire file?

On Fri, 19 Aug 2005 20:44:40 +0200
Denis Corbin <[email protected]> wrote:

> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> 
> yuraukar wrote:
> | My incremental backup created with dar is larger than I expected. The
> | reason seems to be, that I have a large file here, which changes a bit
> | every day -- this is my mailbox.
> | dar includes this file in the incremental backup, but instead of storing
> | only a diff to the original one, it stores the entire file. Is there a
> | way to improve this?
> 
> Not really. If you have a look at diff command, it does not work on
> binary files, it works on text file and replace/add/delete a set of line
> ~ from a file, which is performant if you have only few lines
> chaged/added/removed. But in a binary file there is not that notion of
> line, so diff is useless (try diff on a executable you will see).

Well, there are tools for binary diffs.

> 
> Now what to improve? Making dar detect plain text files from binary
> files would consume a lot of time (need to scan the whole file to know
> if it is plain text file), if so need to analyse the difference. And
> last, need to save the resulting difference or the whole file if it is a
> binary file.

Detection by magic like the file command does isn't particularly time consuming.

> 
> At restoration time, restoring the last version of a given file would
> become painful, as you should need all the archive from the initial full
> backup up to the last differential archive available to restore the file
> in the current state. Actually you only need the archive where has been
> saved the last revision of the requested file.

That is correct.

> 
> What I suggest if you really don't want to save the whole big file that
> changes every day is to make a little script that makes the following:
> Full backup:
> - - make a copy of the mailbox to a special directory
> - - save the copy with dar (or the tool of your choice)
> 
> Differential backup:
> - - make a diff of the current mailbox and the one copied in the special
> directory
> - - save that diff with dar (or the tool of your choice)
> - - replace the copy of your mailbox by the actual mailbox (assuming your
> mailbox has not changed since the diff done two steps above).

And there is IMAP, where the significant data is stored on the server
in single files per message, very good for backups.

Bye
	Racke


-- 
LinuXia Systems => http://www.linuxia.de/
Expert Interchange Consulting and System Administration
ICDEVGROUP => http://www.icdevgroup.org/
Interchange Development Team



-------------------------------------------------------
SF.Net email is Sponsored by the Better Software Conference & EXPO
September 19-22, 2005 * San Francisco, CA * Development Lifecycle Practices
Agile & Plan-Driven Development * Managing Projects & Teams * Testing & QA
Security * Process Improvement & Measurement * http://www.sqe.com/bsce5sf