Is this a timing problem? Machines sometimes boot properly

[Glenn Wilson]

I’m using Fog 1.2 on Ubuntu 13.10 with a proxy DHCP server. My Dell 790 machines will randomly error out (see pic) OR boot into a task. I tried to multicast to 6 machines last night: 3 of them errored out and 3 of them got to the grey download screen (2 had to be rebooted a couple of times because they errored out at first). I am using the undionly.kpxe.INTEL kernel right now.

[url=“/_imported_xf_attachments/1/1597_IMG_20150109_090943343.jpg?:”]IMG_20150109_090943343.jpg[/url]

[Junkhacker]

try using updated ipxe files from here [url]http://sourceforge.net/p/freeghost/code/HEAD/tree/trunk/packages/tftp/[/url]
your problem may already be resolved in trunk

[Glenn Wilson]

I downloaded the latest undionly.kpxe from [url]http://svn.code.sf.net/p/freeghost/code/tags/1.2.0/packages/tftp/[/url] and it SEEMS to be working on both the 790s and 7010s. Both types of machine will now boot to a task or the Fog menu reliably.

BUT there is another issue: When there is an active task, such as an image being deployed to one machine, no other machine, 790 or 7010, will boot past the /default.ipxe… (see pic) part of the process. They will error out right there until the task is completed, and then they will boot to the Fog menu. Any ideas about that?

[url=“/_imported_xf_attachments/1/1598_IMG_20150109_122017650.jpg?:”]IMG_20150109_122017650.jpg[/url]

[Tom Elliott]

Instead of using the [url]http://svn.code.sf.net/p/freeghost/code/tags/1.2.0/packages/tftp[/url]

can you try:
[url]http://svn.code.sf.net/p/freeghost/code/trunk/packages/tftp[/url] files?

[Glenn Wilson]

Tom,

I just tried the undionly.kpxe from the trunk files and it appears to work, except that I had to manually enter the tftp server address before the multicast would start. I’ve seen that with one other kernel I tried this morning. Any idea about that?

[Tom Elliott]

It sounds to me like you may be using DNSMasq?

[Glenn Wilson]

We are using dnsmasq. The /etc/ltsp.conf file shows: “dhcp-boot=undionly.0,xx.xx.xx.xx”.

[Tom Elliott]

I think the middle option can go away to allow dynamic updating of next-server flag:

Maybe the dhcp-boot needs to be:

[code]dhcp-boot=undionly.0,xx.xx.xx.xx[/code]

rather than
[code]dhcp-boot=undionly.0,xx.xx.xx.xx[/code]

The reason the tftp is asked is simply because it doesn’t know what it should be, and would simply fail because it can’t (for whatever reason) find it

[Glenn Wilson]

Do any services need to be restarted after changing the ltsp.conf file?

[Jaymes Driver]

[quote=“Glenn Wilson, post: 40626, member: 240”]Do any services need to be restarted after changing the ltsp.conf file?[/quote]

Yes you need to restart dnsmasq

[code]sudo service dnsmasq restart[/code]

[Glenn Wilson]

I tried the edit and restarting dnsmasq, but the tftp server prompt persists, even after several reboots. Should the change be immediately evident, or does it take some time?

[Tom Elliott]

It would be evident.

I don’t know why it’s not working, but it is why I added the prompt in the first place.

You could create your own undionly.kpxe to thwart this issue by specifying the next-server variable natively and skipping the prompt for the server address entirely.

[Glenn Wilson]

Is it possible to roll my own using the kernel that is working for me now?