Suppose I have two Debian 5.0 servers and a generic session-based DB-driven website is running on them.
Assume that the two DB instances are already clustered.
Each server has its own internal IP address and Load balancing is achieved by putting a Load Balancer in front of the servers (the load balancer listens on the one public IP address assigned to the website).
Now, is there a way to get rid of the load balancer and obtain the same results (with the added benefit of removing the single point of failure) by means of clever network setup (e.g.: logical IP address setup) and the tools that Debian offers ?
If yes, how can that be done ?
Scripts, examples, whatever can aid me in getting there are added bonuses 🙂
EDIT: More specifically, I am looking at something similar to what Windows NLB provides.
Best Answer
To remove the single point of failure, you can do one of the following:
1- Remove the load balancer entirely and configure the servers to work in active-passive mode. You can use heartbeat package for Linux to do it. In this case, you lose the processing power of one machine and use only one as a fail-over!
2- Install and configure another load balancer. The two load balancers will work in active-passive mode. Also, heartbeat can be configured on both servers (I am assuming Linux-based load balancers). This is a bit more expensive as it requires a new system.
3- Install the two load balancers on the same debian systems. So, you don't need to install a new system. However, this may not be recommended as it complicates the systems and mixes more services on the same server (the server already works as web server and DB server).