Thomas Krause wrote: > Hi, > > I tried it with the following rules in /etc/udev/rules.d/19-hd2sd.rules in > my golden client: > > KERNEL=="hda", NAME="sda" > KERNEL=="hda[1-9]", NAME="sda%n" > > With these rules, I can work with my client as if my hard disk is really an > sda (/etc/fstab is working with the sdaX-partitions, etc.). I had only 2 > problems so far: > > 1.) for the kernel root-parameter you still need the original partition > name. To solve this I'm now using disk labels for the kernel parameters as > well as for /etc/fstab. > > 2.) Everything related to partitions in the proc-filesystem is still working > with original real names (obviously). This shouldn't be a big problem > normally, but si_prepareclient for example uses /proc/partitions to detect > partitions in the system and then shows a warning because it cannot find the > devices mentioned there. However the script still continues and worked fine > on my system. To get rid of the warning I explicitly excluded /dev/hda with > the -e option. I guess this won't work if you have partitions that are not > currently mounted, but this is not a problem for me. > > > After this change I still had some problems when I wanted to install the > image on a real client, because SystemImager was now (correctly) expecting > an sda hard disk, but the real clients showed up as hda during the > installation (see previous posts). To fix this I made the same modifications > as above to > /usr/share/systemimager/boot/i386/standard/initrd_template/etc/udev/rules.d/ > * and ran si_mkbootpackage. After that change it worked, but I also noticed > that now the hard disks were already sda even without the udev rules. The > difference to the previous configuration was that I did not use > si_mkbootpackage the first time, but instead copied the kernel+initrd from > /etc/systemimager/boot from my golden client to the /tftpboot-directory. I > don't understand exactly what the difference between these two methods is > and especially why si_mkbootpackage creates a new initrd, if > si_prepareclient already does that, but I'm glad it works now... > > > The udev-rules may still be useful for systemimager, for other sda vs. hda > problems (I saw a few of them in the mailing list) and I like them better > than creating symbolic links as suggested here. And of course they were are > also useful for me to make my virtual golden client more similar to the real > clients, so that I don't need systemconfigurator or other tools to change my > configuration later on. > > Thomas Krause >
Thomas, many thanks for reporting all these interesting and useful info. BTW, maybe a cleaner workaround could be to use only UUIDs in /etc/fstab and create a master autoinstall script using the --autodetect-disk option (both available in si_getimage and si_mkautoinstallscript). I've not tested it, but theoretically in this way we could be totally independent on the particular udev disk naming rules and we should be able to handle the hda -> sda transition problem without any additional configuration... Regards, -Andrea PS which distro did you use for your installations? ------------------------------------------------------------------------- This SF.net email is sponsored by: Splunk Inc. Still grepping through log files to find problems? Stop. Now Search log events and configuration files using AJAX and a browser. Download your FREE copy of Splunk now >> http://get.splunk.com/ _______________________________________________ sisuite-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/sisuite-users
