Re: [Testbed-admins] Problem Customizing generic images

[Robert P Ricci <ricci@cs.utah.edu>]

Thus spake Miguel A. Erazo on Wed, Feb 03, 2010 at 06:00:00PM -0400:
> Hello,
> 
> I added a new node into Emulab by hitting 'Create' in the 'Add Testbed
> Nodes' page. Also, I freed this node already by running the
> 
> nfree emulab-ops hwdown pc1
> 
> All this was performed with the MFSs files that were extracted at /tftpboot,
> until then I DID NOT USE the images that should be extracted at
> /usr/testbed/images. Is this ok?

At the point when you did the 'nfree', the node should have tried to
reload from one of those images. So, you didn't need to use them
explicitly, but if you hadn't upacked them, etc. before you did the
nfree, there could be trouble (which should be easily solvable: just
unpack the images, do 'os_load <node>', then 'nfree' it from whatever
experiment it's sitting in)
 
> I also noticed that when this node boots it is trying to boot from:
> 
> /tftpboot/freebsd.newnode.
> 
> Is this ok? I think it should be booting now from /tftpboot/freebsd

Up until the point when you did the 'nfree', booting from the newnode
MFS should be fine - after that point, it should boot once into the
Frisbee MFS, then reboot and sit in the bootloader until you allocate it
to an experiment.
 
> After doing this I noticed this node was put in the emulab-ops's 'reloading'
> experiment.

This would be normal behavior after doing 'nfree' - we reload the nodes
after each experiment, a
 
> Now, I am customizing the generic image but I got the following error:
> 
> "/libexec/ld-elf.so.1: Shared object "liblthread.so.3" not found, required
> by "imageunzip""
> 
> Any thoughts?

I'm a little confused here - is the node booted off the disk, or off one
of the MFSes?