Shore wrote:

Howdy. I'd like to give Gentoo a try on my newest server. The server isn't doing anything critical at the moment so it's a good choice IMHO. Frankly I'm sick of RH9 and its library nightmare. The box is co-loed about 2 hours away so physical access isn't as easy as I'd like. This eliminates booting from a CD as well. I'm shipping a new SATA drive over to this place next week to install in one of the hot swap bays. I'd like to install Gentoo on it. Is it possible to install Gentoo onto that new drive remotely without booting from a CD and/or bringing down the machine as it is now?


I installed from an existing SuSE installation using chroot onto the new partition,
just following the manual. However, I did have phys. access. The things you should
be aware of especially since you don't are (at least?) the following:
- I would recommend using the vanilla kernel since they tend to be the most stable AFAICT.
- Be very careful where installing kernel modules, and not to overwrite your old
kernel image. Install the new kernel only from within the chrooted environment.
- Be very careful with lilo/grub. Leave the existing options for booting your old system
untouched, and, my recommendation (see below): Do not (yet) change the default
boot image and partition! It is probably best to configure the bootloader from
the *old* installation.
- Install sshd and networking properly.
- When rebooting to Gentoo, use the boot parameter "panic=N", with an
integer N > 0 to tell the kernel to reboot after panicking.
- Also, make sure Gentoo will reboot when anything goes wrong with upping the
network or sshd. It's probably best to use an init-script to reboot after a fixed time, let's say
10 minutes. You then have 10 minutes to see whether ssh works, and whether you can maintain
the Gentoo installation remotely. If you see that it works, you can then remove this
safeguard (possibly with immediate effect in the Gentoo installation). For example,
make an init-script that does "shutdown -r +10 &", which will start a reboot timer
of 10 minutes. If you see that sshd works, you can remotely execute "killall shutdown"
to "defuse the time bomb", and disable the script. Make sure the init script is installed properly
and will actually execute! Copying it into /etc/init.d is not enough! Read the docs.
- If your current kernel has the kexec patch, use it to boot into the Gentoo installation without
changing the bootloader's default. That way, if anything goes wrong, the next reboot will
reboot the previous distro.
- If it doesn't, it may be possible to do something of this kind with lilo rsp. grub alone, but you'll
have to ask someone else. Otherwise, you may be able to patch e.g. lilo so that it boots the old system
in case the time is greater than let's say now + 1 hour. Don't blame me if you don't get this to
work, though! :-)
- If you can't boot to Gentoo without changing the default, boot without the panic option, or with
a very long one to avoid putting the machine into a never-ending booting loop (e.g. panic=3600)
if it does actually fail. And configure your Gentoo init scripts to restore the previous bootloader
configuration after booting. This will require the bootloader config to work properly from the
Gentoo installation (if it worked chrooted, it should work standalone too, however).


Think it through very thoroughly... you have only one try to ensure that Gentoo either comes up
remote-maintainable, or your previous installation is automatically rebooted into, everything
else will require phys. access...


HTH! :-)

(Legal notice: These tips come without implicit or express warranty of any kind. ;-)


-- [EMAIL PROTECTED] mailing list



Reply via email to