1

I'm bringing up LTSP on Ubuntu 14.04 LTS for the first time. The client reaches the point that its console says:

TFTP prefix: /ltsp/i386
Trying to load: pxelinux.cfg/default
Loading vmliuz-3.13.0-24-generic....
Loading initrd.img-3.13.0-24-generic...............ready.

and stops. Depending on the client, it may also say:

Socket failed: Connection timed out or 
Socket failed: Connection failed

then bring up busybox or go into a kernel panic.

How can I debug this?

|improve this question
  • i think it comes from nbd server. perhaps nbd not installed or port not open ? – user276018 Apr 29 '14 at 20:12
  • don't have reputations to provide answer, but I believe this issue can be resolved by opening up your /var/lib/tftpboot/(i386|amd64)/pxelinux.cfg/default and modifying all instances of ipappend 2 to ipappend 3 – awei Apr 23 '15 at 22:05
2

I had same issue and for days couldn't debug it. As last resort I compared working LTSP in Ubuntu 12 with broken LTSP in Ubuntu 14. There was one difference in nbd-server configuration with nbd authentication.

File: /etc/nbd-server/conf.d/ltsp_i386.conf

Nonworking config:

[/opt/ltsp/i386]
exportname = /opt/ltsp/images/i386.img
readonly = true
authfile = /etc/ltsp/nbd-server.allow

Working config:

[/opt/ltsp/i386]
exportname = /opt/ltsp/images/i386.img
readonly = true

Then: service nbd-server restart

It has been working for couple of days without issues.

UPDATE:

The above procedure may not solve your problem (as it turned out, my problems are not solved either). I'd like to share with you some Tips that could help you debugging your problems. Still.. I'm sure that the nbd-server is causing the issue. On the LTSP server do a command: netstat -c or netstat -c | grep nbd . Watch what happens with the nbd connections. When client connects to nbd, the port is being opened (let's say XYZ). The connection gets status ESTABLISHED. Then reboot the client and pay attention at nbd tcp connection. It doesn't close at client shutdown. So when the second time the same client boots, the same nbd port (XYZ) is assigned to the client but that port is already taken by the clients first boot cycle. This is the reason for yout socket failed erros. For NBD specifics read: http://nbd.sourceforge.net/ . I doesn't know why default configuration of nbd-server in Ubuntu 14 acts like this. In Ubuntu 12 default configuration was fine (i have working LTSP deployment on Ubuntu 12 for 8 months now). But theres an alternative of NFS.

NBD differs fundamentaly from NFS. Ubuntu uses NBD, but Debian still uses NFS stating it is more stable and reliable. I have no errors with client booting on Debian LTSP server with NFS.

So my advice to you is either: 1) use LTSP with Debian with default NFS or 2) use LTSP on Ubuntu but switch NBD to NFS.

|improve this answer
  • I just ran into the same issue. I'm stunned that it's caused LTSP to be completely unusable for nearly 2 years and it still isn't fixed. We are now testing the thin client support in PC-BSD. – Aaron C. de Bruyn Jan 18 '16 at 23:53
  • We have the same issue from time to time (Debian 7.7), all clients just stops booting with "socket failed" error. The only workaround we found is to restart nbd-server. – Vladimir M. Jul 20 '16 at 7:10

Not the answer you're looking for? Browse other questions tagged or ask your own question.