also sprach Gabor Gombas <gombasg@digikabel.hu> [2010.01.30.1131 +1300]: > IMHO you are confused: /proc/sys/kernel/domainname is the NIS > domain name and has nothing to do with DNS; in fact, the NIS > domain can be different from the DNS domain. Thus your script is > doubly wrong: on a NIS system hostnames are normally not qualified > so you cannot obtain the NIS domain from the host name, and on > a non-NIS system the NIS domainname should remain unset even if > the hostname is qualified. Okay, so the script is wrong in that it sets the domain name, which can be easily fixed. I don't see how it's "doubly wrong" > Also, mdadm currently works fine with custom-built kernels having > no initramfs at all. Why break that? I am not breaking that, but it's been deprecated for years. In fact, kernel-mode md assembly does not support 1.x metadata — now the default — and that won't be changed. > It would be much nicer to > patch mdadm to try to read /etc/hostname directly if gethostname() > returns an empty string. There is no /etc/hostname in the initrd, but this is an idea. > Regarding to the issue Marco raised: in fact it is quite common to > install a machine on an internal network where it gets some random > name from DHCP, and rename it when it is ready to be put at it's > intended location. When you do not have physical access to the > final location then finding out that the machine no longer boots > with the new name may be painful. It won't be a problem is the mdadm.conf file is in place and accurate. Having the hostname set and correct just means that it would all work even without the conf file. -- .''`. martin f. krafft <madduck@d.o> Related projects: : :' : proud Debian developer http://debiansystem.info `. `'` http://people.debian.org/~madduck http://vcs-pkg.org `- Debian - when you have better things to do than fixing systems
Attachment:
digital_signature_gpg.asc
Description: Digital signature (see http://martin-krafft.net/gpg/)