2. Add PXE booting to DHCP server
Network booting is initiated by a DHCP server sending extra information to a PXE client. At a minimum, we have to tell the booting network device which server will be providing files via TFTP and which network boot program (NBP) to load.
Add the next-server
and filename
directives to the
192.168.16.0
subnet declaration.
allow booting;
allow bootp;
...
subnet 192.168.16.0 netmask 255.255.240.0 {
option routers 192.168.16.1;
option domain-name "hpc";
option domain-name-servers 192.168.16.1;
option subnet-mask 255.255.240.0;
default-lease-time 600;
max-lease-time 7200;
next-server 192.168.16.1;
filename "pxelinux.0";
}
...
You should already have host sections for each compute node in the cluster
network. If not, power down all compute nodes, restart the DHCP server
with PXE boot enabled and listen to your eno4
network device for
DHCP messages. For each compute node, power it on via IPMI and listen
for the PXE network boot DHCP requests. These will transmit the MAC
address you need.
host c01-hpc {
hardware ethernet de:ad:be:ef:09:39;
fixed-address 192.168.17.1;
option host-name "c01";
}
Do not forget to restart the DHCP server every time you modify the dhcpd.conf
file:
[root@master ~]# systemctl restart dhcpd