Webserver with load balancing and ip failover

Aim:

The main aim of the project is to provide a web server as service to clients. Since uninterrupted service is desirable, load balancing and high availability has been implemented as well.

Description:

Infrastructure

There is only one physical machine(running CentOS Linux 7) involved in the whole project. The rest of the infrastructure has been implemented using virtualization.
Libvirt,an open source API has been used for this purpose. It manages KVM/QEMU. Although used quite interchangibly, QEMU is a hypervisor/emulator whereas KVM is a special operating mode of QEMU that uses CPU extensions(HVM) for hardware acceleration, utilising CPU feauture in Intel VT-x or AMD SVM.
But instead of typing in the command for installing(using virsh) manually, which is not very efficient for an environmet which requires working with multiple VMs(virtual machines), the installation process has been automated using vagrant. Vagrant uses Virtual Box as its default hypervisor, but since we are using Linux, we have used Libvirt as its backend. The configuration of the Vagrantfile and the network files(isolated or private network as well as NAT-enabled network, besides management network) have been uploaded in the commits. The commonly used commands in the vagrant environmet are listed:

mkdir tuxlab
cd tuxlab
vagrant init
vi Vagrantfile
vi networkfile.xml

Once the files are ready, we bring up the vitual machines,check its status and login remotely for working.

vagrant up
vagrant status
vagrant ssh vm

For details, visit http://www.vagrantup.org

Network Topology

There are seven virtual machines(indicated by vm1, vm2 and so on). The virtual machines share one common management network(and hence gets an ip in specified range in lab-mgmt.xml file) and an isolated networki(lab-vm-isolate.xml file). Only one virtual machine(vm1 here) has an additional public network that can reach out to the internet. Hence that particular interface(eth2 here) gets an ip address from dhcp server as per the ip range mentioned in the concerned network file(lab-vm-net.xml here). A static ip address(10.10.10.1 here) is set to the eth1 interface of vm1 and all the other VMs get ip in the same network from vm1. IP forwarding and masquerading is enabled in vm1 interface so that the remaining virtual machines can avail internet access via vm1. The configurations for the VMs are:

vm1: DNS server, DHCP server and router
vm2, vm3: cliets
vm4, vm5: web servers
vm6, vm7: load balancers
Services used

The services used and the purposes they meet are listed below:

iptables : To implement iptables rules
dnsmasq : To implement dhcp server
bind : To implement dns server
Nginx : To implement web server
HAProxy : To implement load balancing and high availability
Keepalived : To implement ip failover
Installation and Configuration
iptables Installation: To install and start the service, type in:

yum install iptables-services
systemctl start iptables
systemctl enable iptables
systemctl status iptables

Configuration:
Disable and mask firewalld:

systemctl stop firewalld
systemctl disable firewalld
systemctl mask firewalld

To list all the rules and save them:

iptables -L
iptables-save > /etc/sysconfig/iptables

To configure vm1 as a router, the iptables commands are:

iptables -I FORWARD -i eth2 -o eth1 -j ACCEPT
iptables -I FORWARD -i eth1 -o eth2 -j ACCEPT
iptables -t nat -I POSTROUTING -o eth2 -j MASQUERADE

To enable packet forwarding:

vi /etc/sysctl.conf
net.ipv4.ip_forward = 1
sysctl -p

dnsmasq
Installation: To install and start the service, type in:

yum install dnsmasq
systemctl start dnsmasq
systemctl enable dnsmasq
systmctl status dnsmasq

Configuration:
For the service to listen on its respective port 53, the iptables command is:

iptables -I INPUT -p udp --dport 53 -j ACCEPT
iptables -I INPUT -p tcp --dport 53 -j ACCEPT

The configuration file /etc/dnsmasq.conf is uploaded in the commit. After being done with the configuration, restart the service with the command: systemctl restart dnsmasq

bind
Installation: To install and start the service, type in:

yum install bind
systemctl start named
systemctl enable named
systemctl status named

Configuration:
Open the configuration file and the zone file and edit them:

vi /etc/named.conf
vi /var/named/<domainname>.zone

Open port 67 for the service to listen:

iptables -I INPUT -p udp --dport 67 -j ACCEPT
iptables -I INPUT -p tcp --dport 67 -j ACCEPT

Restart the service: systemctl restart named

Nginx
Installation: To install and start the service, type in:

yum install nginx
systemctl start nginx
systemctl enable nginx
systemctl status nginx

Configuration: For minimal configuration, simply add an index.html file at the default document root of nginx and restart the service

echo "Welcome to web server" > /usr/local/nginx/html/index.html
systemctl restart nginx

For more information, visit http://www.nginx.com

HAProxy
Installation: To install and start the service,type in:

yum install haproxy
systemctl start haproxy
systemctl enable haproxy
systemctl status haproxy

Configuration: Configure the file as uploaded in commit, and restart the service

vi /etc/haproxy/haproxy.cfg
systemctl restart haproxy

For more information, visit http://www.haproxy.org

Keepalived
Installation: To install and start the service,type in:

yum install keepalived
systemctl start keepalived
systemctl enable keepalived
systemctl status keepalived

Configuration: Configure the file as uploaded in commit, and restart the service

vi /etc/keepalived/keepalived.conf
systemctl restart keepalived

For more information, visit http://www.keepalived.org

Conclusion

Testing
To verify the working of the server and test load balancing and ip failover service, a request can be sent from the local host to any one of the proxy servers(using the virtual ip, here 10.10.10.50, assigned by keepalived). While the response can be seen in the client, if one of the web servers or even one of the load balancer service is stopped, even then the client continues to recieve the response.The following script can be used in the client machine to test this:

while sleep 1
do
curl www.tux.lab
done

Hence load balancing and high availability of the web server is done successfully.

poulomihore / web-server-with-load-balancing-and-ip-failover Goto Github PK