Linux – Failover Internet connection

failoverlinuxnetworking

In my Linux server i have three network cards. The eth0 card is connected to ISP1, the eth1 card is connected to LAN and the eth3 card is connected to ISP2.

What i want to do is to automatically use eth3 as Internet connection if the eth0 connection fails. How can this be done?

Another problem is that my firewall is referring to eth0 as the server is doing NAT for clients this would be broken if eth3 is taken in use, how can this be solved?

Best Answer

Here's a link you might find helpful if your network is simple enough: Fallback gateways. For more complicated situations, you'll need solutions like Quagga, an implementation of OSPF protocol for Linux.

Original

From a command prompt:

sudo route add default gw i.p.add.ress eth1

Then, in /etc/network/interfaces, add this line:

up route add default gw i.p.add.ress eth1

Of course, replacing i.p.add.ress with the gateway's IP in both lines.

Edit for new question

If you need to delete a route, just change "add" to "del". Example:

(command) sudo route del default gw 128.0.x.x eth0
(/etc/network/interfaces) up route del default gw 128.0.x.x eth0

You can set a static route for the branch office through the route command as well

route add -net (network_id/prefix) GW (gateway) (interface)

CentOS Dual WAN – Load Balance and Failover

Actually reference you provided, describes a solution for your gateway router. What is misses, it's local lan setup. Here it goes:

set up your NAT, on both interfaces

iptables -t nat -A POSTROUTING -o eth2 -s -j MASQUERADE

iptables -t nat -A POSTROUTING -o eth3 -s -j MASQUERADE

make sure you've enabled forwarding (sysctl -w net.ipv4.ip_forward=1)
check if facebook.com is alive, by pinging it, or trying http request - if it is alive, do nothing ;-) You may assume that if facebook.com is alive, your WAN link works fine. If it's down - people will call you anyway yelling the Internet is down, because facebook does not work anymore. You may write a script to do this.
in case the connection is down - you have to do only one task: add ip rule and direct your traffic through failover WAN link. It primary wan is on again - delete the rule, and flush the cache. Like this:

ip rule add from 1.2.3.0/24 table WAN2

ip route flush cache

or restore traffic:

ip rule delete from 1.2.3.0/24 table WAN2

ip route flush cache

Make your changes permanent. I'm not familiar with CentOS, but sysctl.conf and rc.local should be at your service.
Why you don't have to switch default gateway on your box? According to your reference, you have already added default routes to different routing tables. It means, your box will be still available through secondary routing table. And yes, even with primary gateway dead, you will be able to reach it with secondary table, which has it's own default route.
I'd also avoid balancing traffic through different prociders, as load balancing is based on route cache, so it may cause a lot of unidentified connectivity problems. I'd rather stick with failover.
There's just final thing - if your wan links are private networks, you won't be able to reach your box from outside your network other way than redirecting ports or setting up vpn link. In this case, switching gateways might be required.
For source reference check here.

Best Answer

Related Solutions

How to force Squid Proxy to use a specific gateway for internet access

Original

Edit for new question

CentOS Dual WAN – Load Balance and Failover

Related Topic