Requirement : to build several fujitsu 250 systems with solaris 8 from
a jumpstart server using a flash image.  Client has a script on j/s
server which prompts for solaris version (8), ethernet adress of new
system (gleaned from "banner" in OBP), IP address of new server, new
systemm name, and whether sun or fujitsu build required.  J/S server
(in E450 FWIW) cabled to small hub, new systems also cabled to it.
The Fujitsu servers appear top have three NICs "on board" - a single
one, and below that a pair (one above the other).  The top NIC of each
pair is used to connect to the hub (general advice from another sys
admin here).

In short everything seems set up OK...  the boxes are cabled exactly
the same.  the script on the j/s server places the ethernet address
and hostname in /etc/ethers OK, and the IP address and hostname in the
/etc/hosts file, and an entry in /etc/bootparams.

Problem :  some boxes boot from net and install OK, others don;t even
boot from the net!

ie I run "boot net - install" from the OBP/ok prompt - some systems
bootstrap and immediately do that "number count" thing and progress
through building...  whilst others bootstrap and display one "waiting
for ARP/RARP" message - then do nothing else whatsoever....  no more
ARP/RARP messages (which one would expect if the ethers info say was
wrong), no build starting....  and no response from the keyboard ie
hit return doesn't line feed...  Its as if the very initial bootp
process has begun...  but drops out/freezes/halts/fails/whatever.

This is a closed network in effect, there are no IP clashes.

The systems that fail to respond (other than a single ARP/RARP
message) have been tried also with a successfull system's cable (in
case its the cable or the hub) to no avail, and on both of the paired
NICs (other sys admins say the lone NIC is not working/used/whatever).

Any ideas?  Obviously the jumspstart server is working, the ethernet
address etc is all added correctly, and even the same cables are being
used.

cheers

ian

Check net is correcly set in devalias, also that use-nvramrc? true

It may be that what you are expecting as "net" is not what you are
getting.
Also, check out local-mac-address? you may need that set to false

--
        Bruce

"The internet is a huge and diverse community and  
not every one is friendly"                            
        http://www.ytc1.co.uk                    

- the successfull servers ALL run POST when boot net - install is
entered.  The unsuccessfull ones ALL fail to do so, and go straight to
the "boot" line...

- having snooped the jumpstart server I can see the following traffic
on an unsuccessful system attempt to boot net - install...

cl3      -> BROADCAST   TFTP Read "C0A8F067" (octet)    
jsserver -> cl3         TFTP Error: access violation

(cl3 is the intended hostname of the box to build; C0A8F067 equates to
192.168.240.103, the intended IP address of the box to build)

Googling the above indicates possible problems with the /tftp (or
/tftpboot ?) directory contents...  I have to return to the build area
yet to have a gander at these, but the googling also suggested their
might be an issue with the architecture....   given that ALL the
systems are Fujitsu 250s (and other F250s have been built successfully
using this j/s server...  even I've managed four out of 16!) could
this (combined with the POST thing mentioned above) point to some
OBP/firmware revision differences causing the problems?  This is a
complete quasi-educated guess I accept!

cheers

ian

RARP failures cause a message and rpc.bootparams failures cause a
message.  When you don't get anything, I begin to suspect TFTP problems.

Snoop from the jumpstart server and see what's going on, probably you've
got something wrong in the /tftpboot directory.

--
Darren Dunham                                           ddun...@taos.com
Senior Technical Consultant         TAOS             http://www.**-**.com/
Got some Dr Pepper?                           San Francisco, CA bay area
         < This line left intentionally blank to confuse you. >

From another host on the net, can you tftp to your jsserver and read
that file?  Is the file world readable in the /tftpboot directory?

--
Darren Dunham                                           ddun...@taos.com
Senior Technical Consultant         TAOS             http://www.**-**.com/
Got some Dr Pepper?                           San Francisco, CA bay area
         < This line left intentionally blank to confuse you. >

Red herring, I'd think.

The file needs to exist and be *world* readable.

Casper
--
Expressed in this posting are my opinions.  They are in no way related
to opinions held by my employer, Sun Microsystems.
Statements on Sun products included here are not gospel and may
be fiction rather than truth.

You can use snoop program to see what the messages are coming to the
jumpstart server.  This can tell you whether the client is at least
communicating with the jumpstart server and can rule out network
configuration issues.

If you do not see any communication from the client, it could be the
switch port speed settings. We have best results with auto detect. Could
also be the network interface drivers is not loaded or supported as a
jumpstart interface. we jumpstart off the hme interface. Jumpstart can
NOT be used with gigabit (barried in the docs somewhere).

If you see that the client is trying to communicate via your snoop on
the jumpstart server, then it maybe that the image specificed for the
boot is incorrect. ie, wrong architecture or the OS image on jumpstart
does not have driver support for that server type.  For example to
jumpstart a SUN 240, you need Solaris 8 revision 12/03 on the jumpstart
server. It could be a duplicate for the mac address that has incorrect
information before the correct entry.

Good luck.