cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
818
Views
0
Helpful
3
Replies

Simple failover for 2 webservers using layer 3 cisco switch

umisenken
Level 1
Level 1

I have 2 webservers serving the exact same content and want to provide failover. Both webservers are connected to catalyst L3 switches and I was thinking of using some combination of STP and etherchannels to accomplish this.

Currently we have a software solution which isn't working very well and one of my CCNA friends suggested I try this.

I'm VERY new to Cisco equipment so if any of you could give me a hint on where to go or if I'm doing it wrong that would help a lot.

Load balancing would be a plus but if not we'll stick with Round Robin with a short TTL

Thanks,

Nick

1 Accepted Solution

Accepted Solutions

One option you have is called "Anycast".

The theory is you have multiple "services" with the same IP address, and you advertise that address to a router. If the service goes off line, the route gets pulled.

This is used extensively with DNS servers on the internet. Here's how it works:

Create a loopback interface on the linux boxes (both/all), and assign the address that users use will be using to access the site (same address on both boxes).

Using a routing protocol that can run on the linux boxes (and supports same cost load balancing), advertise that route to the router.

As long as one of the servers is up the route is advertised to the network and valid.

Dan

View solution in original post

3 Replies 3

danrya
Level 1
Level 1

Couple of questions:

Do the web servers serve static content only, or dynamic as well?

What os are the servers running?

Dan

They have dynamic content but it can be treated as static, there are no persistent sessions or anything like that.

I'm running them both on Redhat Enterprise Linux 4

One option you have is called "Anycast".

The theory is you have multiple "services" with the same IP address, and you advertise that address to a router. If the service goes off line, the route gets pulled.

This is used extensively with DNS servers on the internet. Here's how it works:

Create a loopback interface on the linux boxes (both/all), and assign the address that users use will be using to access the site (same address on both boxes).

Using a routing protocol that can run on the linux boxes (and supports same cost load balancing), advertise that route to the router.

As long as one of the servers is up the route is advertised to the network and valid.

Dan