Kernel problems with unionfs and Ubuntu gutsy

[R. F. Grant]

Ok, just to report back that I have now got FAI working to install
gutsy. I did it using the method describe by Tim Cutts, namely:

- Going back to using our existing FAI 3.1.8 server
- Compiling a 2.6.24 etch kernel with all the necessary drivers in it,
and booting from that
- I have to add pci=nommconf to the FAI boot parameters, but when gutsy
is actually installed with the new kernel this is not necessary
- editing /etc/initramfs-tools/initramfs.conf in the nfsroot to boot
from an nfs filesystem
- adding an initrd=.... into the pxelinux.cfg file

It would have been nice to move to fai 3.2.x but it just didn't seem to
work, and nobody who responded seems to have got fai 3.2.x to work with
gutsy with drivers new enough to support the latest hardware (ie. 2.6.22
or later). If anyone ever does manage to get unionfs etc working on the
newer kernels, and manages to install gutsy then that would be
interesting to hear about. But right now, 3.1.8 with custom (non
unionfs) drivers seems to be a good working solution.

Many thanks to all who replied.

Richard Grant,
Senior Computer Officer
School of Mathematics & Computer Science
University of Leicester


Tim Cutts wrote:
>
> On 25 Feb 2008, at 11:07 am, Tim Cutts wrote:
>
>>
>> On 21 Feb 2008, at 3:42 pm, R. F. Grant wrote:
>>
>>> After trying many kernels, I also find an extra annoyance. The new PCs
>>> contain
>>> Intel Q35 motherboards which contain a known bug: they will not boot
>>> from a
>>> regular kernel unless you add pci=nommconf or acpi=off. There is a
>>> patch
>>> that
>>> is supposed to fix this for 2.6.24 kernels, although I have not had
>>> much
>>> luck
>>> with this.
>>
>> I've successfully FAI installed HP Deskpro 7800 desktops, which have
>> a Q35 chipset, using FAI 3.1.8 with a 2.6.24 kernel.  The kicker was
>> the e1000e ethernet driver that these machines need.  Neither of the
>> above kernel options were required with that kernel (this was the
>> 2.6.24-1-686 package from current Debian unstable, rebuilt for etch)
>>
>> I can't address your issues specifically but the above approach
>> worked fine for me.
>
> I should add that the steps I took to make this sid kernel work for
> FAI 3.1.8 were as follows:
>
> 1)  Re-build it on an etch system so that the dependencies were correct
> 2)  Add it to our local debarchiver repository, so that apt can see it
> 3)  chroot into the FAI NFS root on the FAI server and
> 3a) Install the new kernel package
> 3b) rebuild the initrd file to expect the root filesystem to be NFS
> (without this, you get a kernel panic because it can't find the root
> filesystem)
> 4)  Copy the vmlinuz and initrd files out of the NFS root to the tftp
> directory
> 5)  Add an initrd= stanza to the pxelinux.cfg file so that the initrd
> gets loaded when the nodes PXE boot
>
> I imagine some of this is already in place if you use FAI 3.2.4,
> because support for initrd images is designed into that version, isn't
> it?  It's worth noting that installing Debian etch on these HP dc7800p
> machines still requires us to use an additional graphics card in the
> machines; the Q35 graphics chip requires X.org stuff way newer than
> what's in etch, so we just slap in an nVidia card in each machine
> (with the additional advantage that it's DVI; the X packages'
> autoprobing of the monitor resolution seems to work much better that
> way, especially with cheaper widescreen -- 1920x1200 -- monitors, such
> as the Samsung SyncMaster 245B that I have in front of me)
>
> Basically, the Linux support for recent HP deskpros is shaky to say
> the least.  Both dc7700 and dc7800 require very recent kernels if you
> don't want to have to do the acpi=off hack (which has other problems -
> as soon as you switch off ACPI on these machines the interrupt
> handling goes to hell, and the mouse becomes very erratic in its
> responsiveness). I fear this problem is going to become widespread
> though - other cheap business desktops like the Dell 755 are using the
> same chipset now.
>
> I'm still not convinced even with the 2.6.24 kernel that everything is
> absolutely right - in particular the SATA disk performance seems
> pretty poor.
>
> But none of this is relevant to FAI so I'll shut up now.  :-)
>
> Tim
>
>
> --The Wellcome Trust Sanger Institute is operated by Genome
> ResearchLimited, a charity registered in England with number 1021457
> and acompany registered in England with number 2742969, whose
> registeredoffice is 215 Euston Road, London, NW1 2BE.
>