Thanks for all the replies. Too numerous to quote them all.
The answer is no you can't :-(
The original post:
> I am trying to get jumpstart to work over a subnet without a boot server
> on that subnet.
> We have HP Jetdirect cards booting via bootp on the same subnet to a SunOS
> machine (by turning on IP-Helper on the CISCO box)
> The Sun boots fine in the subnet with a boot server in the subnet.
> We are setting up Virtual Lan's and we don't realy want to put boot servers
> all over the place.
> Does anyone know how I can make this work?
> What is the difference bettween Sun's Jumpstart and the bootp that the
> jetdirect cards are using?
> Any ideas?
> Thanks in advance
The replies that helped the most:
You also need a static arp entry for the jumpstart server on the client subnet.
If you do a snoop between the two nodes, you'll see what's going on.
From: "Mark Steph" <firstname.lastname@example.org>
suns dont use bootp. Sun's use rarpd to translate mac address to ip, then
use bootparamd to determine how to boot. What we do at exu is have boot
servers that serve 8 networks each -- with 2 quad ethernet cards. We have
4 of these that allow 1600 clients to boot (at 48 clients per network).
You'll have more trouble than its worth (IMO) with vlans instead of routed
networks. They spent tons of money to convert them to all switched flat
networks here (for everything except the unix networks) and now I think
they see the problems they caused and are going to try to switch them back.
From: Michael Ryan <mike@NetworX.ie>
HP JetDirect uses BOOTP, which is a standard Internet protocol;
Sun JumpStart uses RARP and bootparams. Eventhough RARP is a
standard protocol, bootparams is not. You would need your Cisco to
support bootparams relaying, which, as far as I know, it does not.
Therefore, either get a bootparams relay device or you're out of
--- Received from EEA.EEABHL +61 9 301 9999 97-06-25 11.25
Sorry guys, looks like we can't do it, each subnet requires it's
The way it works is like this:
1) You go off and configure your cisco with an IP helper-address so
that it will forward the RARP request and the boot request from
the diskless Sun to the subnet that has the server on it.
2) You attempt to boot the diskless Sun on the serverless network.
3) Your cisco router forwards the RARP request packet (either
directly to the server, or if you use the more general helper
address, to the broadcast address on the server's subnet).
4) The server looks at the packet, says, "Hum, this guy wants to know
his IP address - Ok, I will tell him." And sends back the client's
5) The client (This is where it all goes down hill) pulls out his IP
address, and the server's MAC address. He then sends the boot
request directly to the server's IP address / MAC address. This
doesn't work, because the router ignores the packet as it is not
addresses to it's MAC address.
Had the client taken the IP address and ARPed for it, the cisco
would have replied, "Sure, I can get it there." Then the client would
have sent the packet to the cisco's MAC address with the server's IP
address and everything would have worked out fine.
/ \ Geoff Lamb, Ericsson Data Australia
| (o)(o) Internet: email@example.com
C .---_) Phone: +61 3 9301 3597
| |.___| Fax: +61 3 9301 3566
Segmentation fault (core dumped)
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:11:58 CDT