VLAN "stretching" over frame relay (WAN in general

Hey mate,

it depends a bit on your IP scheme...
Maybe SLB can answer your question.

SLB: Server-Load balancing.
It depends a bit on the IOS you have running.

I used it a couple of months ago for the same purpose.

The configuration is pretty easy.

What you have to do:

-Configure SLB on your cisco
-Configure a loopback address on your server (same address on both servers)
-Create a DNS record (if you prefer to use DNS) to point to the virtual IP address

Thanks mate...I have updated my diagram to show my IP scheme.

Updated diagram: yfrog.com/3shaclusterdiagj

I did think a bit about SLB but thought it wont solve my problem since the nodes are in active-passive mode. i.e. the node in site A is the active node handling client requests, while the node in site B is in standby mode waiting for the node in site A to fail so that it can take over.

Correct me if I'm wrong but SLB would be ideal if the cluster nodes operated in an active-active mode (i.e. both nodes are active). In other words, a separate node serving as the load-balancer could hold the virtual IP....(not quite sure becos where would this node be located in order to prevent a SPOF? a third site?) I should add that due to the scale of this implementation, both nodes will be located in same location but separated by a frame relay which makes them appear as being in different sites. After this phase then the real implementation with nodes at diff locations will be carried out.

The problem:
As shown in my diagram the virtual IP (an IP alias managed by Heartbeat from linux-ha.org) would not be able to migrate to site B since it's not part of the 192 network so i guess the best solution to have a single network span across both sites. Would it make sense having say a 172 network on both sites?? I doubt it...

Two solutions that come to my mind after research are:

1. Layer 2 extension (dont quite know how to go about it though)
2. Run a routing daemon like Quagga on the node in site B so that it can advertise a route to the virtual IP, leaving upstream routers to deal with routing requests to the node.

Thanks for any more input.

PS. traffic doesnt go via the Internet hence why private IP's have been used..Also feel free to correct my diagram ...

Thanks mate...I have updated my diagram to show my IP scheme.

Correct me if I'm wrong but SLB would be ideal if the cluster nodes operated in an active-active mode...

You're 100% right... I didn't read your post careful enough, sorry.

You can do things like this with content switches such as those sold by F5 Networks. Basically these boxes, which in my opinion shouldn't really be called switches at all, let you do loads of clever front-end things. For your application you would define a vurtual server with a single IP to which your clients would connect. Then behind that you'd configure a server pool containing your two nodes. Whatever failover opccurred within the pool would then be transparent to your users/clients. The one drawback is that these boxes cost money. Price is dependent on how much throughput you're going to need, but you could contact them for a bit of free pre-sales advice and a guide to the likey cost

TheBishop, thanks for the post but I'm restricted to using Cisco hardware.

Thanks again for any inputs.

As far as I know, you can have the same subnet across frame relay. The only thing remaining is how to apply redundancy (i.e automated switching from primary to backup).

It seams that you can do it if you have a "Content Services Switches". Check here:

www.cisco.com/en/US/products/hw/contnetw...186a00801dcd75.shtml

Another way you could try is using 2 DNS servers, one placed at site A and one on B. You configure your clients to use site A's DNS server IP as the primary DNS IP and site B's as the secondary IP. Your clients should connect to the Data Centers using a single DOMAIN NAME (not a direct IP) say datacenter.company.com. Then you would define this domain name/zone on each of the DNS servers but letting them point to their own site data center IP. So DNS of site A will map datacenter.company.com to 172.16.0.1, and DNS of site B will map datacenter.company.com to 192.168.1.1.

Now when site A is normally operating, the primary DNS IP (site A DNS) will be reachable so clients will get 172.16.0.1 when they request datacenter.company.com. when site A is down, the primary DNS IP (site A DNS) will NOT be reachable and clients will be headed to site B's DNS, so they will get 192.168.1.1 when they request datacenter.company.com

Hope this helps.