Fresh install: LTSP on Ubuntu 13.10: "Error: socket failed: connection refused"

[John Hupp]

On 11/30/2013 1:31 PM, John Hupp wrote:
> On a fresh installation of Ubuntu or Lubuntu 13.10 and LTSP, the LTSP 
> client fails to boot. After a good PXE boot, successful TFTP, and the 
> splash screen, boot stops with the message "Error: socket failed: 
> connection refused."
>
> Then I'm dumped to the console at a working initramfs prompt.
>
> I follow with the console messages, starting with the last message 
> that one sees in a good boot, followed by all the bad messages:
>
> Trying to load: pxelinux.cfg/default ok
> mount: mounting /dev/nbd0 on /root failed: Invalid argument
> /scripts/init-bottom/ltsp: line 27: panic: not found
> chroot: can’t execute ‘/usr/bin/test’: No such file or directory
> mount: mounting /root on /rofs failed: Invalid argument
> mount: mounting /rofs on /root/rofs failed: Invalid argument
> mount: mounting /dev /root/dev failed: No such file or directory
> mount: mounting /sys on /root/sys failed: No such file or directory
> mount: mounting /proc on /root/proc failed: No such file or directory
> Target filesystem doesn’t have requested /sbin/init-ltsp.
> No init found. Try passing init= bootarg.
>
> BusyBox v1.20.2 (Ubuntu 1:1.20.0-8.1ubuntu1) built-in shell (ash)
> Enter ‘help’ for a list of built-in commands.
>
> (initramfs) _
>
> -----------------------------------------------
>
> This is on a fresh default installation of Ubuntu 13.10. I followed 
> with this standard LTSP-PNP setup (using a single NIC):
>
> sudo -i
> add-apt-repository --yes ppa:ts.sch.gr
> apt-get update
> apt-get --yes install dnsmasq ltsp-server-standalone ltsp-client 
> ldm-lubuntu-theme
> ltsp-config dnsmasq
> echo 'IPAPPEND=3' >> /etc/ltsp/update-kernels.conf
> /usr/share/ltsp/update-kernels
> ltsp-update-image --cleanup /
> ltsp-config lts.conf
>
> -----------------------------------------------
>
> I modify the installation in just two ways:
>
> 1) To fix LTSP Client Boot Error: "PXE-E32: TFTP open timeout," in 
> /etc/dnsmasq.d/network-manager replace the "bind-interfaces" line with 
> a "bind-dynamic" line.
>
> 2) To fix Internet DNS Name Resolution broken by LTSP setup, edit 
> /etc/dnsmasq.d/ltsp-server-dnsmasq.conf and comment out the port=0 line.
>
> ===============================================
>
> Knowing now that the issue occurs under both Ubuntu and Lubuntu 13.10, 
> it does not appear to be directly connected to the desktop environment 
> or to any changes in the DE between 13.04 (where LTSP worked) and 13.10.
>
> One person suggested that the kernel used in 13.10 might not support 
> my network card, but when booting to the Live CD the network card 
> works fine.
>
> Another person or two suggested that the problem was some issue 
> between the NBD server and client, but so far I have no specific 
> troubleshooting measures. On the server, 'sudo ps -a | grep nbd' shows 
> that there is one instance of nbd-server running. 'Netstat -lt' shows 
> that for tcp (the tcp6 results are irrelevant here, right?), listening 
> is occurring at
> *:9571
> Dell-Ubuntu:domain
> Dell-Ubuntu.loca:domain
> localhost:domain
> *:ssh
> localhost:ipp
>
> Has anyone else experienced this or found out what is wrong?
>


I think I have good evidence now that this is a bug in kernel-level nbd 
support.

Saucy uses kernel 3.11.0-12.19 in its initial default configuration, and 
the error occurs under this kernel.

But I installed 3.9.0-4.9 (which was also published for Saucy), and the 
client boots normally when it uses that kernel.

Testing has been hampered by the fact that some of the pre-3.11.0-12.19 
kernels break networking (e.g. 3.11.0-0.3 and 3.10.0-2.10), and I don't 
know if/how to test for working kernel nbd support except by booting an 
LTSP client. Nonetheless I'll try to narrow down where the failure occurs.