September 7th, 2012, 06:37 AM
Disaster recovery networking solution
Do any network specialists here have a tip on Disaster recovery networking solution?
Which is the best way of having a server in one data centre sharing the same IP with a server in a secondary data centre. If the primary datacentre or server fails the secondary one will take over
September 7th, 2012, 09:53 AM
I could be wrong here, but I'm pretty sure you cannot share an IP address between two data centers. The point of an IP address is to route requests to a specific physical location.
The point of a host name is to route a request to a specific logical location. If you needed to do failover from one datacenter to another you would update your DNS records to point at the IP addresses from the secondary data center rather than the first.
Failing over across datacenters isn't necessarily going to be seamless, which is why they generally have things like backup generators that can run for days or weeks. For seamless failover you need the two servers in the same datacenter; then you can do this at the routing or load balancing level.
September 7th, 2012, 11:23 AM
I heard you can use a BGP router.
Comments on this post
September 7th, 2012, 12:10 PM
The best way is to have a physical Load balancer. What this does is creates a "Virtual IP Address" that's shared for a pool of servers. So let's say the VIP is 192.168.1.5 but the "Real IP"s are 192.168.100.5, 192.168.100.6. You set your DNS entries up for the 192.168.1.5 and the traffic flows to the LB and it then distributes the load based on rules you've defined. E.g. round-robin, last-used, etc.
I forget what the setting is, but there's a registry entry for Windows servers that allows for non-strict naming. Basically there's an issue where if you point to server.test.com and the LB sends you to server01.test.com; the actual "server01" won't recognize the other name even though the connection's come in correctly. Easy fix though and we do it here all the time.