Linux Enterprise Cluster notes-ch10 build a Linux cluster

1. There are not many technical points in this chapter.

2. LVS-NAT cluster. This is a LVS-based load balance cluster architecture. A bunch of servers exit from a NAT server. All requests and returned information must go through the NAT server. Therefore, this Nat Server is a bottleneck and will be detailed later.

3. LVS-DR cluster. dr means "direct routing". In this LVS load balance cluster, all requests go through one server and are scheduled to a bunch of servers. These servers automatically return the response to the client, therefore, it is called direct routing. In this way, the server only stood by the request, the returned information has nothing to do with themselves, so better than the LVS-NAT, this is also the book focuses on a load balance cluster architecture. specifically, this architecture is better than LVS-NAT in the following aspects:

(1) LVS-DR clusters are easier to maintain, because for Server clusters that receive loads, In the LVS-NAT, to maintain these servers, you must first log on to the Load balancer before you can log on to the past; different LVS-DR, you can log on directly to the server group because these servers do not have fake addresses.

(2) The Load balancer server of the LVS-DR only accepts the request, the return has nothing to do with itself, the load is reduced a lot.

(3) In the LVS-NAT, if Load balancer down, the entire cluster has been unable to use; Under the LVS-DR, if Load balancer down, other clusters can still provide services, however, the customer needs to change the IP address of the Request server. I personally think this is far-fetched.

4. If the cluster is monitored. Chapter 1 describes how to use SNMP to monitor clusters.

5. How to monitor cluster performance. Chapter 1 describes how to use ganglia to monitor cluster performance.