Re: [Testbed-admins] Can't get interactive mode while trying to test images

[Ben Burnett <bburnett@atcorp.com>]

ok, that helps alot knowing the "pxeboot.emu-null" means
"non-interactive"  the documentation says it is for no-output when
trying to debug display or serial output problems....  (we'll stop
using it...)

here is a goofy update:
on pc1 "pxeboot.emu-vga" works fine

on pc2 "pxeboot.emu-vga" panics and fails....  so we are trying
"pxeboot.emu-sio"

once we got pc1 up we were able to get booted into FreeBSD62 image
which we are fixing some problems in fstab for disknames and mounting,
but in Linux we get these errors:

testbed config: gethostbyname(UNKNOWN) failed
TMCC exited with status 256!
*** WARNING: Could not get bossinfo from tmcc!
*** /usr/local/etc/emulab/rc/rc.bootsetup:
   Could not determine the name of the boss server!
Boot failure!
gethostbyname(UNKNWON) failed
TMCC exited with status 256!
Error sending boot errno to Emulab Control!
gethostbyname(UNKNOWN) failed
Error sending TBFAILED to Emulab Control!
Boot Complete

Fedora release 8 (Werewolf)
Kernel 2.6.25.14-69.emulab1.fc8 on an i686

pc1.hwdown.emulab-ops.foo.net login:

at least 2 problems:
1. It can't find boss
2. the name is "pc1.hwdown.emulab-ops.foo.net"  it should be
"pc1.hwdown.emulab-ops.emulab.naresky"


what is TMCC?  where is the name and address for boss configured?  is
this supposed to come over pxe?  or is this a custom configuration
step that was left out of the docs?

Thanks,
--Ben






Ben Burnett
  R&D Software Engineer: wired, wireless, ad hoc, mesh networking,
    network monitoring and data visualization
  ATC (Architecture Technology Corporation)   http://www.atcorp.com/
  bburnett@atcorp.com                         (952) 829-5864 x167
  --------Views expressed are mine and not those of ATC--------
  Personal:
    ben@burnett.ws                            http://www.burnett.ws/
http://ben.burnett.ws/
http://www.linkedin.com/in/benburnett



On Tue, Jul 14, 2009 at 1:28 PM, Mike Hibler<mike@flux.utah.edu> wrote:
> I am going to have to go back and figure this all out again sometime...
>
> Okay, so the version of pxeboot you use doesn't affect the output of the
> later boot code (and kernel).  In the case of pxeboot, "null" really means
> "non-interactive."  If you choose the -null pxeboot, you will get output:
>
>  pxe_open: using PXE DHCP info
>  pxe_open: ipaddr: 155.98.39.80/255.255.252.0
>  pxe_open: server addr: 155.98.32.70
>  pxe_open: server path: /tftpboot
>  pxe_open: gateway ip:  155.98.36.1
>  DHCP Info: IP: 155.98.39.80, Server: 155.98.32.70, Gateway: 155.98.36.1
>
> but you will NOT get:
>
>  Type a key for interactive mode (quick, quick!)
>
> and you will get output again when it loads the kernel and MFS.  After
> the kernel loads, what the console is (and thus whether you get output)
> depends on the setting of "console" in /boot/loader.conf (or in the case
> of an MFS, /tftpboot/*/boot/loader.conf.orig).  At this point, a "nullconsole"
> setting really means null and you don't get output.
>
> So, if you want to interact with pxeboot, you need either -vga or -sio.
> After that, if you are not getting the "quick, quick!" prompt, or if it
> doesn't take your input, then we have some other problem.
>
> Sorry for the confusion, it's been a long time since I messed with this code.
>
> On Tue, Jul 14, 2009 at 12:09:49PM -0500, Ben Burnett wrote:
>> we are getting the rest of the output, just not the interactive mode prompt....
>>
>> Leigh, yes this was working before....  before we moved the nodes from
>> the reloading project to hwdown...   they are both in hwdown
>> currently...
>>
>
> What was working before, your ability to interact with pxeboot or something
> else?  I'm trying to figure out what could have changed by going from
> reloading to hwdown using the sched_reserved/nfree combo.  It will reset the
> pxeboot path if you have an alternate.  Do you use different pxeboots for
> different node types?
>
>
>