Cross Site Cluster
What is a Cross Site Cluster
A Cross Site Cluster, when used in conjunction with other tools, such as Cross Site Private Network, allows users to create clusters of servers or load balancers that span across multiple network zones. This provides load balancing between the zones which can be used to provide an automatic failover in the event of an outage.
In this configuration, all servers are allocated the same additional IP addresses. Only the servers in Reading will receive all of the traffic under normal conditions. If a server begins to fail its health checks, traffic to this will be directed to the other server in Reading.
If both servers in Reading fail their health checks, or there is a data centre wide outage, traffic will be redirected automatically to the standby server in Dunsfold.
Ideally, in the above scenario, 4 IP addresses would be allocated to the servers, and only one of them would be “preferenced” for each server. This would mean that under normal operation, traffic for a particular IP address would only be routed to one server. However, if the health checks of this server fails, the traffic for this IP address will be re-routed over to one or more of the other members of the cluster. Round-robin DNS can then be used to distribute traffic to all of these IP addresses, and hence between all members of the cluster across both sites.
There are some constraints that must be considered for the Cross Site Cluster service:
- Servers and load balancers must be on separate firewalls. This may mean that in order to add some members to the cluster, they may need to be moved to a different host/firewall. Our support team can advise on this and it may result in some downtime.
- Servers should be on a different primary vLAN. This is because we are unable to ensure that traffic is directed to the correct host, as any host on the same vLAN with the IP configured can respond.
- Servers shouldn't have a matching preference for an IP address as any other server in the same network zone. This is because traffic will only ever be routed to one of these servers, even if the health checks are passing for both. In this configuration, a health check failure on one server will still cause the traffic to be re-routed to a different server with the same preference.
Last updated 10 October 2017, 09:11 GMT