etherboot-users

[Etherboot-users] DHCP- Etherboot not sending DHCPREQUEST

[David T. P. <phe...@un...>]

Hello,

    I am trying to install Rocks 4.1 Fuji on a 25 node LinuxNewtorx=20
cluster that was donated to my college. Each node has only a power,
 serial console, and network connections. There is no cdrom or floppy =
drive,=20
so PXE booting these nodes is my only method of installing an OS on=20
these machines.

    The nodes have a LinuxBIOS and Etherboot5.0.4eb4.1 burned onto the =
prom.
My problem is that my nodes are not sending a DHCPREQUEST
to the frontend (DHCP) server.

    I have worked with DHCP a number of times but never had this=20
problem. However, this is my first time trying to use PXE booting. Here =
is the output of `tail /var/log/messages` when I power on the first =
node:

Nov  7 12:36:41 manganese dhcpd: DHCPDISCOVER from 00:30:48:22:b7:3f via =

eth0
Nov  7 12:36:41 manganese dhcpd: DHCPOFFER on 192.168.1.253 to=20
00:30:48:22:b7:3f via eth0
Nov  7 12:36:41 manganese dhcpd: DHCPDISCOVER from 00:30:48:22:b7:3f via =

eth0
Nov  7 12:36:41 manganese dhcpd: DHCPOFFER on 192.168.1.253 to=20
00:30:48:22:b7:3f via eth0

The node itself shows:

=FCROM segment 0xc800 length 0x4000 reloc 0x9400
Etherboot 5.0.4eb4.1 (GPL) Tagged ELF (Multiboot) for [EEPRO100]
Boot from (N)etwork or from (L)ocal? N
Found Intel EtherExpressPro100 at 0XC000, ROM address 0X0000
Probing...[EEPRO100]Ethernet addr: 00:30:48:22:B7:3F
Searching for server (DHCP)...
....................No Server found
<abort>
Boot from (N)etwork or from (L)ocal?

    The cluster is on its own network, with a single HP ProCurve 4108GL=20
switch. I searched the Rocks mailinglist archives to see if this problem =

has occurred before, and came across a thread relating my
troubles to the Spanning Tree settings of a SMC switch. I have reset my=20
switch to factory default, re-configured it and made sure the Spanning=20
Tree Edge was set to 'Yes'. When that didn't work, I tried a 12 port=20
unmanaged switch and still had my DHCP problem.
If I connect a laptop to the network, it receives a DHCP address=20
correctly, as does the DHCP client of the switch (for connecting to it=20
and managing it). So, I KNOW dhcp works. I just can't figure out why=20
certain clients will have a successful DHCP transaction and my nodes =
won't.

Any ideas?

Thank you in advance,

--David

--=20



****************************
****************************
**  David T. Pheasant     **
**  Computer Engineering  **
**  Union College 2006    **
**                        **
**  phe...@un...    **
****************************
****************************

Re: [Etherboot-users] DHCP- Etherboot not sending DHCPREQUEST

[Anselm M. H. <an...@ho...>]

Am Dienstag, den 08.11.2005, 16:50 -0500 schrieb David T. Pheasant:
> Hello,
>=20
>     I am trying to install Rocks 4.1 Fuji on a 25 node LinuxNewtorx=20
> cluster that was donated to my college. Each node has only a power,
>  serial console, and network connections. There is no cdrom or floppy
> drive,=20
> so PXE booting these nodes is my only method of installing an OS on=20
> these machines.
> =20
>     The nodes have a LinuxBIOS and Etherboot5.0.4eb4.1 burned onto the
> prom.
> My problem is that my nodes are not sending a DHCPREQUEST
> to the frontend (DHCP) server.
>=20
>     I have worked with DHCP a number of times but never had this=20
> problem. However, this is my first time trying to use PXE booting.
> Here is the output of `tail /var/log/messages` when I power on the
> first node:
>=20
> Nov  7 12:36:41 manganese dhcpd: DHCPDISCOVER from 00:30:48:22:b7:3f
> via=20
> eth0
> Nov  7 12:36:41 manganese dhcpd: DHCPOFFER on 192.168.1.253 to=20
> 00:30:48:22:b7:3f via eth0
> Nov  7 12:36:41 manganese dhcpd: DHCPDISCOVER from 00:30:48:22:b7:3f
> via=20
> eth0
> Nov  7 12:36:41 manganese dhcpd: DHCPOFFER on 192.168.1.253 to=20
> 00:30:48:22:b7:3f via eth0
>=20
> The node itself shows:
>=20
> =FCROM segment 0xc800 length 0x4000 reloc 0x9400
> Etherboot 5.0.4eb4.1 (GPL) Tagged ELF (Multiboot) for [EEPRO100]
> Boot from (N)etwork or from (L)ocal? N
> Found Intel EtherExpressPro100 at 0XC000, ROM address 0X0000
> Probing...[EEPRO100]Ethernet addr: 00:30:48:22:B7:3F
> Searching for server (DHCP)...
> ....................No Server found
> <abort>
> Boot from (N)etwork or from (L)ocal?

Just an idea... you did set the "next-server" option in dhcpd.conf,
didn't you? Etherboot relies on its presence, which had been there by
default for ages, but the most recent dhcpd will not send it out if not
explicitely configured to do so. Try entering
	next-server 192.168.1.1;
in your dhcpd.conf and see if that helps Etherboot to find a DHCP
server.

Hth
Anselm

Re: [Etherboot-users] DHCP- Etherboot not sending DHCPREQUEST

[David T. P. <phe...@un...>]

Just checked the dhcpd.conf and the entry for etherboot pxe is as follows=
:
 else if substring (option  vendor-class-identifier, 0, 9)  =3D "Etherboo=
t" {
                # Chroot TFTP appropriately"
                filename  "pxelinux.0";
                option vendor-class-identifier  "PXEClient";
                option PXE.mtftp-ip 0.0.0.0;
                vendor-option-space PXE;
                next-server 192.168.1.1;
        }

A professor of mine relayed my problem to a friend of his who responded w=
ith
this intresting bit of info:

Client is probably not ACK'ing the DHCPOFFER because the OFFER is
missing something (http://www.ltsp.org/documentation/pxe.howto.html)
shows "option vendor-encapsulated-options
3c:09:45:74:68:65:72:62:6f:6f:74:ff;" for etherboot clients.  It's
quite ingenious really, etherboot can be designed to only accept
responces with that vendor string so that you can have a cluster
DHCP server not conflict with a normal DHCP server.  Of course now
without that option Etherboot clients will not accept OFFERS from
normal dhcpd setups.

As I mentioned in my original message, this is a LinuxNetwox Evolocity
cluster and in my opinion they would do something sneaky like this to kee=
p
customers from switching to an OS/software other than theirs. Is there
anyone that has such a cluster and could send me their dhcpd.conf file?

-David

----- Original Message -----=20
From: "Anselm Martin Hoffmeister" <an...@ho...>
To: "David T. Pheasant" <phe...@un...>
Cc: <eth...@li...>
Sent: Wednesday, November 09, 2005 4:55 AM
Subject: Re: [Etherboot-users] DHCP- Etherboot not sending DHCPREQUEST


> Am Dienstag, den 08.11.2005, 16:50 -0500 schrieb David T. Pheasant:
>> Hello,
>>
>>     I am trying to install Rocks 4.1 Fuji on a 25 node LinuxNewtorx
>> cluster that was donated to my college. Each node has only a power,
>>  serial console, and network connections. There is no cdrom or floppy
>> drive,
>> so PXE booting these nodes is my only method of installing an OS on
>> these machines.
>>
>>     The nodes have a LinuxBIOS and Etherboot5.0.4eb4.1 burned onto the
>> prom.
>> My problem is that my nodes are not sending a DHCPREQUEST
>> to the frontend (DHCP) server.
>>
>>     I have worked with DHCP a number of times but never had this
>> problem. However, this is my first time trying to use PXE booting.
>> Here is the output of `tail /var/log/messages` when I power on the
>> first node:
>>
>> Nov  7 12:36:41 manganese dhcpd: DHCPDISCOVER from 00:30:48:22:b7:3f
>> via
>> eth0
>> Nov  7 12:36:41 manganese dhcpd: DHCPOFFER on 192.168.1.253 to
>> 00:30:48:22:b7:3f via eth0
>> Nov  7 12:36:41 manganese dhcpd: DHCPDISCOVER from 00:30:48:22:b7:3f
>> via
>> eth0
>> Nov  7 12:36:41 manganese dhcpd: DHCPOFFER on 192.168.1.253 to
>> 00:30:48:22:b7:3f via eth0
>>
>> The node itself shows:
>>
>> =FCROM segment 0xc800 length 0x4000 reloc 0x9400
>> Etherboot 5.0.4eb4.1 (GPL) Tagged ELF (Multiboot) for [EEPRO100]
>> Boot from (N)etwork or from (L)ocal? N
>> Found Intel EtherExpressPro100 at 0XC000, ROM address 0X0000
>> Probing...[EEPRO100]Ethernet addr: 00:30:48:22:B7:3F
>> Searching for server (DHCP)...
>> ....................No Server found
>> <abort>
>> Boot from (N)etwork or from (L)ocal?
>
> Just an idea... you did set the "next-server" option in dhcpd.conf,
> didn't you? Etherboot relies on its presence, which had been there by
> default for ages, but the most recent dhcpd will not send it out if not
> explicitely configured to do so. Try entering
> next-server 192.168.1.1;
> in your dhcpd.conf and see if that helps Etherboot to find a DHCP
> server.
>
> Hth
> Anselm
>
>
>
>=20