2007/7/9, Pavel Nemec <pnemec@suse.cz>:
We boot standard system (small kernel with small initrd) from PXE/DHCP and than download "working" image over ftp. Then we unpack this working image and chroot in it. All configurations are done off-line, when image is prepared. Nice! I didn't know about that project. Redhat had a similar project called "statless linux". The main idea was to separate program and configuration files which are static, from dynamic system files (like in /var and in /tmp) and user files. This way they could mount root filesystem read only and one image could be shared among many clients. And one can install it either on hard disk, on CD(DVD), USB key or mount it as a NFS share.
My personal opinion is that separation like this should be in the mainstream distribution in the first place. It makes cloning a lot easier. But dynamic files (/etc/mtab is the first example that comes to mind) kreept into parts of filesystem which are not meant to be dynamic. Files like this make liveCD-s and diskless clients dificult to prepare and update. Tools like unionfs would not be necessary if there was clear separation. System images made with KIWI are targeted for read-only media, so the separation of static and dynamic files is already there. That is why I find it suitable for system cloning. Of course for instalation on hard disk one does not need compresion and instead of building compressed image one could just make root filesystem tree and untar it. You could easily update the image and cloning would then just consist of taring the / again (and maybe /home and /srv). Wiki page on cloning seems very good Idea to me. Martin -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org