When I worked at Cisco in a former life, we had ~300 redhat machines that we completely installed and managed remotely. When we laid out the filesystems we also created a second root partition. The purpose was to have a minimal OS that would allow you to boot the machine and re-install the main os on the other partitions. To make it work we entered 2 entries in /etc/lilo.conf, one for the real / and one for the recovery /. in the rc.local on the main /, we added a "lilo -R main" to do a one time boot of the main /. If that boot failed, due to bad fsck'ing, or anything else that prevented the machine from booting all the way, a re-boot would bring the machine up on the known good alternate /. To make sure this alternate / was not corrupted, we would either never mount it under normal running mode, or mount it read only. To actually get the machine installed, we created a boot floppy that booted the machine and mounted / over nfs. the machine then had the above mentioned alternate / created, and installed with the minimal OS. The machine would then re-boot on to the alternate / and then install the main system. To get around the floppy disk problems, we checked if there was a disk inserted, and if there was, we updated lilo on the floppy to boot the proper root (nfs or alternate / or main...) We could then send someone a dd of the floppy disk, and rawrite, and have them make a boot disk for us, and insert it into the machine. Once that was done, no other user intervention was required. In all the time I was there we only lost 2-3 machines that would not come back up with the above setup. If you want to look at the script that created the boot disk and did the os install, you can find it at http://cvs.sourceforge.net/cgi-bin/viewcvs.cgi/ceps/scripts/newinstall.sh You may need a few other of the utilities there to actually get it to work properly. There are probably better ways to accomplish this, we started this back with redhat 5.0, or so. So the install isn't a kickstart install, but it shouldn't be too hard to integrate it into a kickstart config. -Adam Jim Barbour wrote: > > Hello, > > I suspect that my problem has been seen before, so I'm curious how > others have dealt with it. > > We will need to install linux on many machines in multiple data > centers, and would very much like to do it remotely. > > I plan to configure everything to use the serial port, so I believe > I'll be able to control the installs remotely. However, there's the > matter of removing that pesky boot floppy. > > I saw something in the syslinux documentation that (sort of) implied > that I might be able to configure syslinux to boot off a kernel on a > hard drive, so I could setup a "continue" label that says (go boot off > the linux kernel on /dev/hd0a) > > I'm wondering if anyone can provide guidance on getting syslinux to > boot a linux kernel using a .BS file, or if anyone has other ideas > about how to solve this remote installation issue. > > Again, thanks for your time, and your help... > -- > Jim Barbour --- Staff Engineer, Systems Programmer/Administrator > ------------------------------------------------------------------------------- > Office: 858-651-2616 "Standing next to me in this lonely crowd > Pager: 800-200-6068 There's a man who swears he's not to blame. > Cel: 619-977-6491 Every night I hear him shout so loud, > Alphapg:jbarbour@pager Cryin' out that he's been framed." - Bob Dylan > ------------------------------------------------------------------------------- > > _______________________________________________ > Kickstart-list mailing list > Kickstart-list@xxxxxxxxxx > https://listman.redhat.com/mailman/listinfo/kickstart-list
