On 1/28/19 8:14 AM, Song BJ Yang wrote:
The network boot device specified in BIOS, eth2 in your example, will send out dhcp discover packet, however, since nodders.installnic is set to eth1, hence the management server(DHCP server) side only presents dhcp lease for "eth1", "eth2" cannot obtain lease from management server(if no dhcp range configured),the network boot will fail, and the next boot device specified in bios "boot order" will attempt.

Thanks for your answer, but I still don't get it : I'm not talking about site:dhcpinterfaces attribute ? Are you saying that nodres.installnic is server side related ?

I might have been unclear :

What I'm saying is that my client nodes have

eth0 : 1G connected interface
eth1 : 1G disconnected/unplugged interface
eth2 : 10G connected interface

and are configured in BIOS to PXE boot first on the "XE" card, which matches the 10G/eth2 interface

Besides it is the port this card is plugged into which I'm using (successfuly) for swith-based discovery.

a) as a matter of fact, I can observe that all traffic, from initial PXE to discovery to osimage download occurs via this "eth2" network card on the client.

BUT, it may look strange since

b) the "noderes.installnic" attribute is set to <my management host>|eth1",,

which I'm not sure about what it is meant to (and I've read the man page) :

So the only way I can think of how this works is that xNBA is first instructed to try to PXE on eth1 (because of noderes.installnic == eth1 and then, since PXE cannot work (unplugged interface), fallsback to eth2.
If so, again how does it choose between eth2 and eth0 ?


[ Side note : on the xCAT management server, the "site.dhcpinterfaces" attribute is set to eth1 which is the interface facing the same network as the nodes, which is of course expected]

Thanks for you help

--
Thomas HUMMEL



_______________________________________________
xCAT-user mailing list
xCAT-user@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/xcat-user

Reply via email to