On the loadbalancers:

On the web servers:

General configuration is only within /etc/keepalived/keepalived.conf. All the following configuration elements should be placed in this file (by default, it contains an example configuration).

First, the module ip_vs needs to be loaded at boot. Add it to /etc/modprobe.preload. Normally, keepalived will load this module automatically.

#modprobe ip_vs

Keepalived is be used to manage ipvsadm as well as the fail-over between the loadbalancers. We have established redundancy between the loadbalancers for security reasons, to make sure we don't have to worry about this. Keepalived also installs ipvsadm, the ip_vs administration tool.

Let's make sure that it implements ip_forward, because it will not function without it:

echo "1" > /proc/sys/net/ipv4/ip_forward

Edit /etc/keepalived/keepalived.conf. The global_defs section applies to the general configuration of Keepalived. It allows you to define notification options, and mainly the lvs_id (unique machine identifier).

global_defs {
    notification_email {
    mon@mail.com
    }
    notification_email_from keepalived@domaine.tld
    smtp_server 127.0.0.1
    smtp_connect_timeout 30
    lvs_id LVS_MAIN
    }

The vrrp\_instance section defines use parameters of the VRRP daemon included in Keepalived. This protocol handles dynamic IP addressing between many systems. This allows redundancy of the loadbalancers. If one becomes unavailable, the other picks up the load and grabs the IP address specified in virtual\_ipaddress{}. For the second loadbalancer, the state variable contains the SLAVE value. virtual_ipaddress defines all the IP address that the loadbalancers should attribute itself when it is the master. Be careful, virtual_ipaddress only handles one IP per line with a maximum of 32 addresses.

vrrp_instance VI_1 {
    state MASTER 
    interface eth0
    virtual_router_id 51
    priority 100
    advert_int 1
    authentication {
    auth_type PASS
    auth_pass 1111
    }
    virtual_ipaddress {
    10.0.0.5
    }
    }

The following section discusses defining virtual servers for redundancy, as well as the protocols to implement it.

# Virtual Servers definitions
    
    virtual_server 192.168.0.1 80 {
    delay_loop 30
    lb_algo wlc
    lb_kind NAT
    persistence_timeout 50
    protocol TCP

If all the real servers are unavailable, keepalived will return the request. It allows you to show a static information page in case of a problem.

sorry_server 192.168.0.1 80

Add all the real servers:

real_server 192.168.0.12 80 {
    weight 20
    }
    
    real_server 192.168.0.13 80 {
    weight 8
    }

    }

Since the machines are within a local network with the loadbalancers, we must force them to use the loadbalancers as gateways for all traffic. Let's create a static route to do this with iproute2.

Let's create the following rules:

#!/bin/sh -e
    echo 200 LVS >> /etc/iproute2/rt_tables
    /sbin/ip rule add from 192.168.0.12 table LVS
    /sbin/ip route add default via 192.168.0.1 dev eth0 table LVS
    /sbin/ip route flush cache

Let's test the configuration. First, restart keepalived:

# service keepalived restart

Use your favorite Web browser to access http://10.0.0.5. If the page loads correctly, your configuration is working. Of course, this is a basic configuration. It may be interesting to add more advanced features such as firewalling, monitoring, etc.

Once the server is installed, you can use ipvsadm -L to view the state of your load balancing:

# ipvsadm
    IP Virtual Server version 1.2.0 (size=4096)
    Prot LocalAddress:Port Scheduler Flags
    -> RemoteAddress:Port           Forward Weight ActiveConn
    InActConn TCP  10.0.0.5:www wlc persistent 20
    -> 10.0.0.3:www                  Route   25     294        916       
    -> 10.0.0.4:www                 Route   8      243        754       

Keepalived includes a monitoring system capable of checking if the servers in your high availability cluster are reachable. You only need to add a section in the Web server's configuration in /etc/keepalived/keepalived.conf. The monitoring can be a simple ping:

TCP_CHECK { connect_port 80 connect_timeout 3
    }

It can go as far as checking that the returned page is the page it should be.

HTTP_GET {
    url {
    path /
    digest 69c520033af926d40e7d5d832e3933f6
    }
    connect_timeout 5
    nb_get_retry 3
    delay_before_retry 2
    }

To manage the hash, use genhash with the following command:

[root@lb1 ~]# genhash -s 193.188.255.4 -p 80 -u /
    MD5SUM = 5b5af20a92f1dbe92328005329d1753f

You can also launch actions if a specific event happens (typically, fall over to the other loadbalancer if the master becomes unreachable), with the following set up in section vrrp :

notify_master /path_to_script/script_master.sh
    (or notify_master “/path_to_script/script_master.sh <arg_list>”)
    notify_backup /path_to_script/script_backup.sh
    (or notify_backup “/path_to_script/script_backup.sh <arg_list>”)
    notify_fault /path_to_script/script_fault.sh
    (or notify_fault “/path_to_script/script_fault.sh <arg_list>”)