[Testbed-admins] problem (?) with emulab-ops/hwdown

[Chaos Golubitsky <chaos@bbn.com>]

Continuing my tour through the FAQ: i want to boot and modify one
of my images, upgrading the Emulab software per [1].  To do that,
i want (i think) a node booted from a copy of the image on its
disk, with access to NFS-mounted testbed filesystems.

According to [2], a good way to reserve a node for maintenance is:

  boss> nalloc emulab-ops hwdown pc1

I assumed that i could then reboot the node into the MFS, use
imageunzip to copy the bits i want onto the disk, and the reboot
and ask for "part:1" or what-have-you.  I hoped that i'd get a
booted node in the emulab-ops/hwdown experiment, and, in particular,
that i'd have access to /proj/emulab-ops and other nice things.

However, when i try to boot that partition, i see the boot-time
error:

  Jun  3 12:25:29 pc1 evproxy[816]: could not get IP address from hostname: pc1.hwdown.emulab-ops.pgeni.gpolab.bbn.com
  *** /usr/local/etc/emulab/rc/rc.bootsetup:
      Error running /usr/local/etc/emulab/evproxy
  Boot failure!

I can login as root, but i don't have any NFS-mounted partitions,
and presumably no other setup either.

Indeed, if i do that lookup on boss:

  boss> host pc1.hwdown.emulab-ops.pgeni.gpolab.bbn.com
  Host pc1.hwdown.emulab-ops.pgeni.gpolab.bbn.com not found: 3(NXDOMAIN)

Any thoughts here?  In particular, is the problem most likely to be:
 (A) nalloc is incorrectly failing to setup DNS mappings for hwdown
 (B) the old image i'm trying to upgrade is incorrectly insisting
     on DNS mappings
 (C) the hwdown experiment isn't the right way to get a booted node
     which can mount /proj/emulab-ops
 (D) the DNS issue is a red herring; the NFS failure is caused by
     something else
 (E) something else entirely

Thanks again.

Chaos

[1] http://users.emulab.net/trac/emulab/wiki/UpdateNode
[1] http://users.emulab.net/trac/emulab/wiki/kb116