nfsboot ubuntu desktop fails with errors creating /dev/null

[Twigathy]

I installed Ubuntu Fesity desktop with the aim of creating an nfsboot system, the files being stored on a server on the LAN. The client machine (mostly) comes up, I do get a login prompt for a very short period of time before more services decide to load and then the error regarding /dev/null appears.

I checked on the server side, /dev/null does exist in the client image and seems to have the correct permissions.

I do not really know if this is a bug or an error with my config; I'm fairly confident my setup is correct...

Stuff done:
(On the client)
Created /mnt/edam
Mounted 192.168.0.15:/mnt/storage/NFS -> /mnt/edam
cp -ax /. /mnt/edam/. (Created the NFS boot image)

[edit boot image on 192.168.0.15]
Edited fstab - remove entry for /
      - Add entries for proc and sys

Edited /etc/network/interfaces
      - Removed all entries except loopback

[add entry to /boot on the VM]
cd /etc/initramfstools
vim initramfs.conf (Change from "local" to "nfs")
update-initramfs
vim initramfs.conf (Change back to "local")
mv new.initramfs to newramfs-NFS
mv new.initramfs.bak to newramfs
add entry to grub.lst

[Try it and see :o]
 - still problems with /dev/ (Can't create /dev/null(!))
- Errors after the /dev/null one:
/etc/init.d/rc: 2: sed: Input/output error. (Repeated for a few screens).

[Robin (subscriptions-robinandmariette)]

I'm having the same trouble except that I don't see the /dev/null error, although it may flash by too fast for me to read. I get a slew of complaints about sed before the system hangs.

I added "single" to the boot options of my nfs bootup, and successfully got to a root prompt. The only problem I had during the boot (that I could catch) was a timeout trying to mount my samba share. From there I could do pretty much anything that I wanted. The network was up, I had internet access, I could nfs share other file systems, I could start X and get a full desktop etc.

Once I verifiied everything was looking good, I went back to the root prompt and entered "telinit 2". From there it was all bad. Avahi timed out, so I removed that from the startup and tried again, and next the "network manager" (i think it was called) failed after that. Once that happened I got a flash of messages, and the last screen was full of startup script failures complaining they couldn't find "sed" (it's in /bin and does exist on the system). Since the system hangs at that point I can only guess what went wrong. I assume something is blowing away my root file system.

Next I'll start manually running the rc2.d scripts and see if I can narrow down what's causing all the fuss.

[Twigathy (twigathy)]

It does look to me like some networking components in bootup are killing it off. It's 2:15am where I am, so I may well poke this tomorrow! :)

[Twigathy (twigathy)]

I did a quick search on launchpad; it looks like we are not alone!
https://bugs.launchpad.net/ubuntu/+source/avahi/+bug/111227

There seem to be a couple of "solutions" listed, but they are a bit hacky and nasty....

[Twigathy (twigathy)]

Hacked up "/etc/init.d/{avahi-daemon, dbus} with "exit 0" at the top of both files to get them to go away and stop mucking around with the network card when it already has an IP.
Testing....Fails. New errors!

Seems to fail almost immediately after "Starting basic networking..."
Ends up with basically file not found errors and then I/O errors. Which suggests "Starting basic networking" is hosing eth0 on boot.

I can't work out which service is responsible for this message though. Must be somewhere in /etc/init.d/ :-)

Seem to be getting a tiny bit closer...

[Twigathy (twigathy)]

Here's a screenshot showing current progress (With nasty hacks).
A few seconds later the screen fills with I/O errors.

[Robin (subscriptions-robinandmariette)]

It's NetworkManager, and it's binary /usr/sbin/NetworkManager, not a script, so it will be hard to patch. I get:

NetworkManager: <information> Deactivating device eht0
NetworkManager: <information> Will activate wired connection eht0 because it now has a link

[Twigathy (twigathy)]

Hey,

I managed to get it all working pretty nicely. I've written up what I did here:

https://help.ubuntu.com/community/Installation/OnNFSDriveWithLocalBoot

[Nanley Chery (nanoman)]

Did you figure out if this was an error on your part, or if it was an actual bug?

[Twigathy (twigathy)]

Hi,

Yeah, turns out this was actually because of Network Manager taking down and bringing up again eth0. That "bug" is still present.

My feisty (NFS root) install has both these files contents as "exit" to stop network mangager mangling the network connection:
/etc/default/NetworkManager
/etc/default/NetworkManagerDispatcher

[Nanley Chery (nanoman)]

Thank you for taking the time to report this bug and helping to make Ubuntu better. The issue that you reported is one that should be reproducible with the live environment of the Desktop CD of the development release - Lucid Lynx. It would help us greatly if you could test with it so we can work on getting it fixed in the next release of Ubuntu. You can find out more about the development release at http://www.ubuntu.com/testing/. Thanks again and we appreciate your help.