a half-dog high and a dog-and-a-half long

i've been looking at leasing dedicated servers on which to run openbsd, and have only found a few that offer openbsd as an operating system choice. quite a few more support freebsd, and many support at least one flavor of linux. some refuse to install openbsd, or charge a hefty service fee to have someone manually install it.

with dedicated servers available for just $60/month, it's unfortunate that the operating system has anything to do with the service. you are just paying for bandwidth, power, and hardware, but lack of openbsd support ("support" simply being a one-time installation taking all of 15 minutes) severly narrows the list of available vendors, and thus, equipment, price, network, bandwidth, location, etc.

once the operating system is installed, the hosting provider never touches the server and doesn't care how you change the software. a "linux" dedicated server doesn't necessarily have to remain running linux once the provider hands it off, and the provider would probably never even know anything was changed.

so i had an idea: would it be possible to lease a "linux" dedicated server and somehow bootstrap an unattended openbsd installation once it's running without being able to use physical installation media or change any bios boot options?

if one had a working linux server online and knew the disk size(s), ip address, routing information, and other hardware specifics, could a custom bsd.rd image be created that, when booted, would run through the openbsd installation automatically and partition the hard drive depending on its known size, use the existing ip information to do an ftp installation, and then reboot afterwards to come back online as a functioning openbsd system?

or perhaps it would be easier to fit sshd on a ramdisk image that would boot, configure its network interface with the known ip address, and just start sshd to just let one ssh in and walk through the normal installation procedure remotely.

i believe that if a bsd.rd image was just dd'd to the beginning of the hard drive while linux was running and then the machine was rebooted without linux doing any sync()'ing, that the machine could reboot and the ramdisk would boot the next time around.

of course you'd only get one shot at it, and if the reboot, subsequent ramdisk boot, or installation failed, you'd be left with a dead machine and would have to have the provider re-install linux.