Help very confused by diskless node wiki page

[LIsLinuxIsSogood]

I just read the page about diskless node and something that doesn’t seem that clear from the initial portion about the kernel, is where do I download it and is it architecture specific in terms of my master/slave kernels for the PXE boot systems. You see, I am planning on making use of older laptop for a diskless node that I can run with various bootable systems. I am fairly comfortable with all the steps that ARE THERE in the Wiki page such as the networking configuration, dhcp server, and finishing the pxe configuration.

But since the two system have two different architectures, with the master being a 64-bit (or technically with multi lib profile). and the slave system is one 32-bit which I can get to boot from the minimal installation cd.

In terms of the kernel install for PXE where would this come from, is its download? Or can I just build it on my master machine running different architecture? I am confused.

My understanding was that kernel like most other files that are stored specific to the architecture of the machine and that means I either need a working kernel (like the one on the installation cd or I need to find a way to compile a new kernel for the slave somehow). Or can I skip this step altogether by finding a working 32-bit kernel from online?

UPDATE: I guess I had skipped past the section dealing with architectures, because I see it now with respect to the kernel source directory, which is pretty much what I was asking. It looks like I can just copy the bzImage from the correct architecture, so hopefully that works.

[LIsLinuxIsSogood]

So following my lengthy post here is a simple question that I was hoping to be told if this makes sense to do.

I want to recompile the kernel module-less, does that mean I can just replace all M with Y in the config and then run make again?

[NeddySeagoon]

LIsLinuxIsSogood,

Try it ... then you tell us.

The only complication is if you need out of kernel modules like nvidia or vboxdriver.
You will need to leave module loading support on and build those things against your kernel.

A word of warning. 64 bit and 32 bit kernels use the same source tree.
Building a 32 bit kernel for your diskless node will overwrite all the 64 bit build for the host and vice versa.
You need to keep the .config files outside the kernel tree and better yet, keep two separate kernel trees.
_________________
Regards,

NeddySeagoon

Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail.

[P.Kosunen]

[szatox]

You need 32 bit image for your 32 bit pxe client.
The server architecture is irrelevant, none of its code runs on your client.

Regarding "replace M with Y", It did the trick for me, however I already had a pretty well tuned kernel to start with. It's more likely to fail with generic setup than it was with mine, where I disabled quite a few things I knew I wouldn't need.

[WareChameleon]

[NeddySeagoon]

WareChameleon,

With Gentoo its generally a bad thing to go straight to automated deployment without 'learning the ropes' required to maintain Gentoo first.

Over the years we have seen users here making a mess of a update on an install they didn't understand because they skipped the learning/understanding that comes with doing the install.
Worse, these installs were full of user data that had to be saved too, so the option to throw it away and start again was not available, unlike the early days of an install.
Most but not all were salvaged. A few impatient users decided that Gentoo was not for them.

Maybe you want a 'Health Warning' at the start.
Portage has a deliberately poorly documented environment variable that can be used to override some of the safety checks.

As always, its your gentoo your way. When it breaks, you can keep the pieces.
_________________
Regards,

NeddySeagoon

Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail.

[WareChameleon]

I suppose it could be.

The main reason I am putting these scripts out there is because they worked really well and could be used to learn the process.

These days I am working on electronics so I wanted them to continue to live if possible, especially if they help people.

This script doesn't play god with Portage though.

It merely takes a template Terminal and rips through the boring steps you'd normally have to do manually at least once with the Diskless Tutorial on the Wiki anyway.

So once you have one Terminal with everything tailored, you can deploy a school or a library in 10 minutes basically.

Once you decide to build another version, you can copy the files and use the redeploy command to refresh all Terminals.

[Goverp]