no conlusion all people on this list agree to.
you can melt down the discusssion to the following:
1.)linux-ext2-dump is guaranteed to work correct if you have an
completely inactive and sync'ed filesystem. in other words:
if your fs is not mount at all or at least mounted read-only.
in all other situations there is a risk of having file-data and
filesystem-metadata in buffers of the kernel and not on disk.
As linux-ext2-dump avcesses the raw device to get it's images
it bypasses these buffers, and therefor you can/will loose data.
so linux-ext2-dump is not designed to backup active filesystems.
The result of doing that may varry from a completly fine backup-image
over partially garbled files up to a completely useless image,
without getting told about that, as dump thinks all went ok.
And you are right, in my oppionion this should be mentioned at
least in the manpage of ext2-dump in big bold letters.
So i concluded for me not to use ext2-dump for my systems.
but inbetween:
you noticed the version-number of ext2-dump beeing 0.4.xx ?
did you?
this states it as BETA software in develeopment status, not yet
ready to public production-use.
2.) Gnu-tar reads files via the normal filesystem-calls,
and for that has not to worry about unwritten buffers, as it
always gets the correct data from the kernel.
This means you can use gnu-tar for active filesystems, with a few caveats:
a.) some files may get truncatetd/deleted while tar is reading them.
tar will whine and whistle about it, telling you the file
changed while reading it, but your image will be
completely readable, with these files beeing not correct.
b.)gnu-tar changes the accesstime of every file it backups,
so if you need this atime you won't be able to use tar.
i hope this sheds some light on the problem.
Christoph
PS: this is my conlusion to this topic, if your's differs from
mine, please don't flame me. deciding to use ext2-dump on active
filesystems is left to you. i won't perue you from not doing it,
i simply try to show the facts.
Uncle George wrote:
> does this mean that there was a definitive conclusion?
>
> Christoph Scheeder wrote:
>
>>Please not again this discussion...
>>
>
