At time of writing, on a standard Kimsufi, OVH provide you with one static IP, and three “failover IP’s” these “failover” IP’s are just normal IP’s, although they don’t have their own gateway (since only the host machine is allowed to use the gateway). The only thing that’s different about these “failover” IP’s is that if you have more than one Kimsufi or dedicated server with OVH then you can simply change which server the failover IP is routed to, which is quite handy for DR (disaster recovery) or even load balancing.

Let’s get started, first, you need to set up your server with the default CentOS image that is provided by OVH (this may already be the case if you requested this image when you ordered), if not, this is done by selecting your server from the dropdown box at the top of the manager screen, then clicking on “Services” beneath the “Dedicated servers” menu on the left-hand side. While in this menu,  ensure that “Netboot” is set to “hd” so that the boot process is kicked to the hard disk and not OVH’s custom kernels. Next, choose “Reinstall / Change OS” and follow the steps here to install CentOS, the partitioning doesn’t really matter, where possible just choose the default options and continue through the steps until the installation process starts.

The installation process will continue and wipe your server clean and replace it with OVH’s CentOS image (which I might add, isn’t a TRUE CentOS kernel, as provided by the CentOS project). Once you’ve received your email from OVH confirming that the reinstall is complete and have got your new root password, log into the server via SSH. When you have a prompt, type the following commands:

mkdir /cleaninstall
cd /cleaninstall

For 32-bit CentOS:

wget http://ftp.hosteurope.de/mirror/centos.org/5/os/i386/images/pxeboot/initrd.img
wget http://ftp.hosteurope.de/mirror/centos.org/5/os/i386/images/pxeboot/vmlinuz

For 64-bit CentOS:

wget http://ftp.hosteurope.de/mirror/centos.org/5/os/x86_64/images/pxeboot/initrd.img
wget http://ftp.hosteurope.de/mirror/centos.org/5/os/x86_64/images/pxeboot/vmlinuz

Then copy the pxe boot files to the /boot partition

cp vmlinuz /boot/vmlinuz.cent.pxe
cp initrd.img /boot/initrd.img.cent.pxe

Next we need to set up the GRUB bootloader:

yum install grub

Empty the bootloader sequence and insert our new config:

cat /dev/null > /boot/grub/menu.lst
vi /boot/grub/menu.lst

Paste the below text in this file amending the following variables to suit your info:
YOURPASSWORD = A password of your choosing, this is a a temporary password and is only used during the installation via VNC
IPADDR = The IP address of your OVH server (host)
GATEWAY = The gateway IP address of your OVH server (find this by doing: # netstat -r | grep default)

For 32-bit CentOS:

default 0
timeout 30
title Centos Install (PXE)
        root (hd0,0)
        kernel /boot/vmlinuz.cent.pxe vnc vncpassword=YOURPASSWORD headless ip=IPADDR netmask=255.255.255.0 gateway=GATEWAY dns=213.186.33.99 ksdevice=eth0 method=http://ftp.hosteurope.de/mirror/centos.org/5/os/i386/ lang=en_US keymap=us
        initrd /boot/initrd.img.cent.pxe

For 64-bit CentOS:

default 0
timeout 30
title Centos Install (PXE)
        root (hd0,0)
        kernel /boot/vmlinuz.cent.pxe vnc vncpassword=YOURPASSWORD headless ip=IPADDR netmask=255.255.255.0 gateway=GATEWAY dns=213.186.33.99 ksdevice=eth0 method=http://ftp.hosteurope.de/mirror/centos.org/5/os/x86_64/ lang=en_US keymap=us
        initrd /boot/initrd.img.cent.pxe

Save the file (in this sequence hit Escape, colon (:), x, Enter)

Next we need to install GRUB into the MBR of the server disk so…

grub-install /dev/sda
grub-install --recheck /dev/sda

Once that’s done, we need to bounce the box:

shutdown -r now

Wait a couple of minutes for the server to come back online (you can monitor it using ping requests). Once the server is contactable again fire up a VNC client and in the connection box type in the IP address of your server followed by “:1″, for example, if your IP was 192.168.0.5, the VNC connection box will read 192.168.0.5:1

Enter the password you specified as and you will be presented a VNC window at the first step of the CentOS installation screen. Follow the installer (make sure you leave the network options alone – for now at least). After the installation completes and reboots, you now have a PROPER CentOS installation free from OVH’s meddling hands.

Install VMWare Server 2 as you would normally on CentOS, there are tons of HOWTO’s out there on that subject I’m sure. One thing I will say about installing VMWare Server 2 is that when asked about configuring a bridged network interface, to be honest, I wouldn’t configure one, you’ll only need NAT and Host-Only networking. To configure a Bridged interface just invites a headache of having your switch-port blocked if you ever switch on a guest VM which tries to make use of it – which can be very frustrating and WILL cause an interruption to service while OVH sort out unblocking your MAC address from their network.

So, once we’ve got VMware Server 2 installed and running, we need to set up the networking so that the “failover IPs” can be used for the virtual guests. We’re going to be doing this using routed IP’s as it’s the simplest way, without having to involve NAT’ing (ugly).

We’ll start off by configuring the host machine to allow proxied arp requests on the VMware host-only interface, which in this case is the “vmnet1″ device – this is the default interface name that VMware Server creates, however if yours is different you’ll need to amend the below steps accordingly (this is only if you’ve had a non-standard VMware install).

echo "net.ipv4.conf.vmnet1.proxy_arp=1" >> /etc/sysctl.conf && sysctl -p

Next on the host, we have to create a static route to allow traffic to flow to the “failover IP’s”

IPFAILOVER is to be subsituted with one of your failover IP’s. e.g 10.0.0.11

ip route add IPFAILOVER dev vmnet1

A line like the one above is required for each failover IP which you’ll be routing.

This would not survive a reboot (bad) so we have to do something a little more inventive…
We need to create a network config file in the /etc/sysconfig/network-scripts folder to set up the interface dynamically, since VMware doesn’t actually configure the interface using the standard method. In short, we need a config file to set up the interface so that CentOS is “aware” of it, and then we also need to create a route file for the interface too to set up the routing for the interface at boot.

So…
First off we need to find out what interface subnet the vmware installer picked to use for the vmnet1 interface on the host server, so do a

ifconfig vmnet1 | grep "inet addr"

From the output of the above command, use the “inet addr:” address for the IPADDR= and use the “Mask:” for the NETMASK=

Create and populate the /etc/sysconfig/network-scripts/ifcfg-vmnet1 file with the below info:

vi /etc/sysconfig/network-scripts/ifcfg-vmnet1

DEVICE=vmnet1
IPADDR=*.*.*.*
NETMASK=*.*.*.*
ONBOOT=yes

Now that CentOS is aware of that network interface, the routing config to the VM’s fail-over IP needs to go into /etc/sysconfig/network-scripts/route-vmnet1 which is done by:

vi /etc/sysconfig/network-scripts/route-vmnet1

IPFAILOVER dev vmnet1

A new line like the one above is required for each failover IP which you’ll be routing.

Now, unfortunately, that doesn’t solve all our problems because VMware server starts up after the networking service and thus after the virtual device is created. So we need a way to restart the networking service after VMware has started. The quickest and dirtiest way of doing that is by putting something in /etc/rc.local which will do the job. I hated having to do this, purely because it’s not “kosher” with the distro, and in general bad practise, but needs must…

echo "/sbin/service network restart &> /tmp/network-restart" >> /etc/rc.local

Ok, so now we need to allow traffic through to the virtual machines. This is done using the iptables kernel module on the host machine (or can be done on the guest so long as there’s a rule on the host allowing all traffic through to that failover IP). Generally speaking, I like to keep my firewall rules in one place for ease of management, and that’s on the host. So, we’ll open up the iptables config file (or use my script to make it easier):

vim /etc/sysconfig/iptables

Find the line:

-A RH-Firewall-1-INPUT -m state --state NEW -m udp -p udp --dport 1194 -j ACCEPT

and insert this line below it subsituting where necessary (below example is allowing port 22 (ssh) access through to this failover IP):

-A RH-Firewall-1-INPUT -m state --state NEW -m tcp -p tcp --dport 22 -d FAILOVERIP -j ACCEPT

As above, a new line (rule) like the one above is required for each failover IP which you’ll be routing.

As far as the server, or rather, host side is concerned, we’re finished. Now we just need to configure the guest OS with the failover IP address (or addresses if you so choose). The virtual machine should be configured to use the following:

Linux hosts:
IP= FAILOVERIP
NETMASK=255.255.255.255
GATEWAY= FAILOVERIP (yes it’s the same as the interface IP)

Windows hosts:
IP= FAILOVERIP
NETMASK=255.255.255.248
GATEWAY= FAILOVERIP (yes it’s the same as the interface IP)

The DNS nameservers can be set anything you wish to use, so long as they’re addressable and accessible by the VM’s. I tend to use OpenDNS‘s servers unless I’m using my own.

It reads “i > u” and in case you don’t “get it” – nevermind. That’s the beauty of it

#!/bin/bash
TESTING_MINS=2
vim /etc/sysconfig/iptables
clear
QUESTION1="Do you want to restart the firewall now? (hit 't' to test for $TESTING_MINS min(s)) [y/n/t] "
echo -n $QUESTION1

a=""
while test -z "$a"
do
        read -n1 a
        echo ""

 case "$a" in
  Y|y)
        echo -e "Restarting...\n\n"
		/sbin/service iptables restart
  ;;
  N|n)
        exit 0
  ;;
  T|t)
        echo -e "Time is now `date +%H:%M` -firewall service will be stopped at `date +%H:%M -d "+$TESTING_MINS min"`\nIf your test was successful, you will need to manually start the service again by running:\nservice iptables start"
        echo "/sbin/service iptables stop &> /dev/null" | at now + $TESTING_MINS min &> /dev/null
		echo ""
        /sbin/service iptables restart
  ;;
  *)
        a=""
        echo -n $QUESTION1
  ;;
  esac
done

P.S. Any scripts I write and publish here are © Rob Freeman and released under the GPL unless otherwise stated.