===== What? =====
My notes on installing RHEL on iscsi. Here iscsi target is Fedora18, RHEL is installed in KVM guests. iscsi can be used in 2 ways:

  * directly in KVM. That way livbirt/qemu do the magic, the guest only "sees" a usual disk.
  * having anaconda booting, initiating the connection to the target and installing on it
  * https://jpmens.net/2011/07/18/network-booting-machines-over-http/

===== iscsi RHEL deployment on KVM =====
<code>
# The following is used to create the virtual machine
virt-install -n $HOSTNAME -r 1024 --cpu host --nodisks \
        --location /mnt/store/deploy/repos/$DISTRO/ \
        --vcpus 2 -w bridge=virbr0 --os-variant rhel5.4 --graphics vnc \
        --extra-args "ip=$IPADDR netmask=255.255.255.0 ks=http://192.168.4.1/ks/$HOSTNAME repo=http://192.168.4.1/x86_64/$DISTRO"

# the following snippets in the kickstart file do
# - configure the installation on the iscsi target
# - create a simple single partition, so the vmlinuz/initrd can be accessed easily after installation
# - also /etc/grub.conf can be used easily
part / --size 256 --asprimary --grow --fstype=ext4
iscsiname iqn.2009-11.com.example:storage.disk1
iscsi --ipaddr=192.168.4.1 --port=3260 

With this the system can be installed on the iSCSI disk.
</code>

===== booting the iSCSI RHEL with explicit ip handover =====
<code>
# After installation we will mount the iscsi target to access kernel/initrd and /etc/grub.conf
# Here the iscsi target backend is blockdevice /dev/loop8, so we create the partition devnode
# and mount it
kpartx -ay /dev/loop8
mount /dev/mapper/loop8p1 /mnt/tmp
$ ls -1 /mnt/tmp/boot/initramfs* /mnt/tmp/boot/vmlinuz* /mnt/tmp/etc/grub.conf
/mnt/tmp/boot/initramfs-2.6.32-358.el6.x86_64.img
/mnt/tmp/boot/vmlinuz-2.6.32-358.el6.x86_64
/mnt/tmp/etc/grub.conf

cat /mnt/tmp/etc/grub.conf


virsh edit rhel5u8i
# and modify for this:
  <os>
    <type arch='x86_64' machine='pc-1.2'>hvm</type>
    <kernel>/mnt/tmp/boot/vmlinuz-2.6.32-358.el6.x86_64</kernel>
    <initrd>/mnt/tmp/boot/initramfs-2.6.32-358.el6.x86_64.img</initrd>
    <cmdline>ro root=UUID=3a8ab4f8-b383-4517-bc9e-3eee06704b02 ifname=eth0:52:54:00:03:0e:c2 netroot=iscsi:@192.168.4.1::3260::iqn.2009-11.com.example:storage.disk1 rd_NO_LUKS rd_NO_LVM ip=192.168.4.22::192.168.4.1:255.255.255.0:rhel6u3i:eth0:none iscsi_initiator=iqn.2009-11.com.example:storage.disk1 rd_NO_MD SYSFONT=latarcyrheb-sun16 crashkernel=auto  KEYBOARDTYPE=pc KEYTABLE=us rd_NO_DM LANG=en_US.UTF-8</cmdline>
    <boot dev='hd'/>
  </os>
</code>

===== booting the iSCSI RHEL, including setting of iBFT =====
  * get ipxe.org ISO from https://ipxe.org
  * boot it, interrupt with <strg>+<b>, on the ipxe prompt:
<code>
ifstat
dhcp
sanboot iscsi:192.168.4.1:::1:iqn.2009-11.com.example:storage.disk1
</code>
  * now we need to change the kernel options for iBFT. After boot of the system also change in ifcfg-eth0:
    * "ro root=/dev/mapper/vg_rhel6u3i-lv_root rd_NO_LUKS ip=ibft iscsi_firmware=1 ifname=eth0:52:54:00:1d:f9:f9 rd_NO_MD rd_LVM_LV=vg_rhel6u3i/lv_root rd_LVM_LV=vg_rhel6u3i/lv_swap rd_NO_DM LANG=en_US.UTF-8"
  * after boot of the system use "BOOTPROTO=ibft" in /etc/sysconfig/network-scripts/ifcfg-eth0

===== booting the iSCSI RHEL, including setting of iBFT, using VLAN =====
  * get ipxe.org ISO from https://ipxe.org - current builds include the required patches
  * direct ip config (did not work for me, only dhcp): https://ipxe.org/cmd/set
  * boot it, interrupt with <strg>+<b>, on the ipxe prompt:
<code>
ifstat
vcreate --tag 100 net0
dhcp net0-100
sanboot iscsi:192.168.4.1:::1:iqn.2003-01.org.x:disk1
</code>
  * alternatively, static ip:
<code>
vcreate --tag 123 net1
set net1-123/ip 192.168.4.23
set net1-123/gateway 192.168.4.1
set net1-123/netmask 255.255.255.0
ifopen net1-123
sanboot iscsi:192.168.4.2:::1:iqn.2003-01.org.x:disk1
</code>
  * for now the initrd does not evaluate VLAN info from the iBFT. You can modify the initrd to pick up the VLAN from iBFT.
  * booting from harddisk: %%sanboot --no-describe --drive 0x80%%
  * https://ipxe.org/cmd/sanboot
===== reading iBFT infos on booted linux =====
<code>
$ modprobe iscsi_ibft
$ iscsiadm -m fw
# BEGIN RECORD 6.2.0-873.2.el6
iface.initiatorname = iqn.2010-04.org.ipxe:05f0f0cb-29f4-3f30-8792-f3bf89b3426f
iface.transport_name = tcp
iface.hwaddress = 52:54:00:1d:f9:f9
iface.bootproto = STATIC
iface.ipaddress = 192.168.4.237
iface.subnet_mask = 255.255.255.0
iface.primary_dns = 192.168.4.1
iface.vlan_id = 0
iface.net_ifacename = eth0
node.name = iqn.2009-11.com.example:storage.disk1
node.conn[0].address = 192.168.4.1
node.conn[0].port = 3260
node.boot_lun = 01000000
# END RECORD
</code>
  * kernel: ip=ibft
  * sanhook
  * 'man dracut' for all commandline options. Can these options use ibft instead of details?