xris_xcess
September 25th, 2008, 08:07 AM
I've decided to setup a diskless environment in the school I work in. Im currently running a Hardy x86 server and Hardy x86 clients.
My pxe booting, netinstall and diskless work, but I've come up against the problem of different hardware on the clients.
I realize that usually when you deploy a computer lab, you tend to setup identical hardware on a number of machines. Unfortunately this is not my case (sort of).
I've got a number of similar machines: Sempron CPU, 1gb ram, Nvidia chipset, nvidia 6100 GeForce video. Now, I setup my system on one of them just as I want to serve it up from the server. Passed it over to the server. Setup the clients to boot into PXE and startup from the server.
First problem. I can't by any means get my nvidia driver/resolution settings to "stick" properly.
When I boot up my setup machine, all is OK. But when I boot a second "similar" machine... the settings are off. Now, I'm not bothered too much about using the restricted drivers (no gaming or heavy graphics here) but I WILL be extremely picky about the resolution.
It seems that the auto configure feature in Hardy does not work too well. I tried "skipping" this by configuring xorg.conf directly, but it doesn't work. Most of my monitors work at 1024x768x60Hz. When I get one machine running properly, I boot up another one... but I'm back at 800x600.
Ideas?
Could I:
1-Create settings for groups of similar/identical machines to be applied as they boot (before gdm and X start).
2-A Script to identify the hardware and write out a xorg.conf for that machine AND that session of X (so as not to interfere with other machines that have already booted).
3-Some how tweak the hardy auto-configure routine for the resolutions that I want (1024x768 in most cases, with some machines at 800x600)... and hopefully load the nvidia driver for those machines that can use it.
4-Give up, it's too hard with a mixed setup.
by the way, I followed the guide at https://help.ubuntu.com/community/DisklessUbuntuHowto.
I guess, instead of doing a nfsroot, I could build a squashfs image for groups of machines and just share the /home over nfs , and save some space.
Anyone with some experience with this kind of setup?
My pxe booting, netinstall and diskless work, but I've come up against the problem of different hardware on the clients.
I realize that usually when you deploy a computer lab, you tend to setup identical hardware on a number of machines. Unfortunately this is not my case (sort of).
I've got a number of similar machines: Sempron CPU, 1gb ram, Nvidia chipset, nvidia 6100 GeForce video. Now, I setup my system on one of them just as I want to serve it up from the server. Passed it over to the server. Setup the clients to boot into PXE and startup from the server.
First problem. I can't by any means get my nvidia driver/resolution settings to "stick" properly.
When I boot up my setup machine, all is OK. But when I boot a second "similar" machine... the settings are off. Now, I'm not bothered too much about using the restricted drivers (no gaming or heavy graphics here) but I WILL be extremely picky about the resolution.
It seems that the auto configure feature in Hardy does not work too well. I tried "skipping" this by configuring xorg.conf directly, but it doesn't work. Most of my monitors work at 1024x768x60Hz. When I get one machine running properly, I boot up another one... but I'm back at 800x600.
Ideas?
Could I:
1-Create settings for groups of similar/identical machines to be applied as they boot (before gdm and X start).
2-A Script to identify the hardware and write out a xorg.conf for that machine AND that session of X (so as not to interfere with other machines that have already booted).
3-Some how tweak the hardy auto-configure routine for the resolutions that I want (1024x768 in most cases, with some machines at 800x600)... and hopefully load the nvidia driver for those machines that can use it.
4-Give up, it's too hard with a mixed setup.
by the way, I followed the guide at https://help.ubuntu.com/community/DisklessUbuntuHowto.
I guess, instead of doing a nfsroot, I could build a squashfs image for groups of machines and just share the /home over nfs , and save some space.
Anyone with some experience with this kind of setup?