I have many low end atom servers for a startup project, each server have a few public ip addresses (no private network).
I'm currently in the process of making it failsafe. There now is a load balancer in front of my http servers.
This load balancer is at this point a single point of failure, when it goes down everything becomes unreachable.
I was wondering if there is a low cost solution to remove the single point of failure?
I looked into round robin DNS, but it is quite unreliable.
In addition I looked into heartbeat which looks like a nice solution when you have access to the hardware and can connect network cables.
But in my case i can't attach network cables or add hardware. I only have root access and can install software and reconfigure Linux.
Update
Thanks for the answers, i appreciate it.
But when using heartbeat or Keepalived you need 2 network cards in your server, my server only have 1. That one network card is the main one.
I can't use the main one because when it gets high traffic the heartbeat communication may fail (and trigger shutdown). I don't want that 🙂
Is there a alternative?
Best Answer
I use Linux Virtual Server http://www.linuxvirtualserver.org/ along with Keepalived http://www.keepalived.org/ to have 2 redundant / fail-over load-balancers.