Question concerning automated Gentoo installation

OnTheRun · n00b Joined: 25 Nov 2003 Posts: 36

Hi all,
I've found this: http://www.scms.waikato.ac.nz/~baz/ document and I'm working trough it.
Has anyone done this before?
Because I did most of the tutorial, but now I don't get any further. My problem is, that I don't know, what I have too boot on the client.
Do I have to make a boot floppy with syslinux (which contains the gentoo-kernel and runs auto_install.sh) or how do I get the client to start the auto_install.sh script?

Thanks and with kindly regards

Maedhros · Posted: Fri Jan 14, 2005 11:18 am Post subject:

From the looks of it, the clients boot over the network from the server you've configured - so you need to configure them to boot from the network card, rather than a floppy.
_________________
No-one's more important than the earthworm.

OnTheRun · n00b Joined: 25 Nov 2003 Posts: 36

Oh ok,
but in the document a pxe resp. syslinux gets setup to boot from. Why this then?

Thanks and with kindly regards

Maedhros · Posted: Fri Jan 14, 2005 11:53 am Post subject:

Erm, I don't quite know what that sentence means... :oops:

From what I can work out from reading the article, you set up a server which contains the installation files, and some scripts for the installation. When the client boots, it looks for a kernel over the network, which is the PXE kernel you set up on the server. When that finishes booting, the auto_install.sh script is run. (This seems to be a bit out of date now, and you'll probably need to change some of the versions it wants to download...). When that's asked you a few questions, it will mount the binary package repositary from the server, and then chroot. In the chroot, another script is run, which installs Gentoo on the local computer using the binary packages from the server.

The tutorial doesn't seem to go into much detail about what happens after the install, though, so I don't know exactly how the client boots after the install has finished. I'd assume you would change the bios and make it boot from the local disk, though. The whole thing seems to be written from the point-of-view of what happens on the server, and it seems you have to already know how to make the clients work somehow.

Hope this makes things a bit clearer.

_________________
No-one's more important than the earthworm.

OnTheRun · n00b Joined: 25 Nov 2003 Posts: 36

Oh, thank you very much. It is muuch clearer now!
How do I get the client to boot over network to that address?
And how can I emerge the whole portage as binary (something with emerge -b) ? Do I need the whole portage tree?

Thank you and with kindly regards

Maedhros · Posted: Fri Jan 14, 2005 2:26 pm Post subject:

The client will will automatically do a dhcp broadcast to find the server it should boot from, so you won't have to set that up.

As for the binaries, the part under "Creating the binaries" in chapter 5 sort of explains that - it suggests manually installing with -b on one of the machines, so that all the files you actually need will be built, and then copy them from /usr/portage/packages into /mnt/cdrom which is a directory from the server shared over the network.

It also talks about a metadata.idx file, but I have no idea what one of them is, or how to create one.

_________________
No-one's more important than the earthworm.

OnTheRun · n00b Joined: 25 Nov 2003 Posts: 36

Thank you very much for your help!
So the network-booting is required for the client?
And um which packets do I need for standard (for example without X)?

Thanks and with kindly regards