【正文】 摘自 ： 《 Linux 企業(yè)集群》 (The Linux Enterprise Cluster) 英文原文 : The Linux Enterprise Cluster Overview This chapter will introduce the cluster loadbalancing software called IP Virtual Server (IPVS). The IPVS software is a collection of kernel patches that were merged into the stock version of the Linux kernel starting with version . When bined with the kernel39。s routing and packetfiltering capabilities (discussed in Chapter 2) the IPVSenabled kernel lets you turn any puter running Linux into a cluster load balancer. Together, the IPVSenabled cluster load balancer and the cluster nodes are called a Linux Virtual Server (LVS). The LVS cluster load balancer accepts all ining client puter requests for services and decides which cluster node should reply to each request. The load balancer is sometimes called an LVS Director or simply a Director. In this book the terms LVS Director, Director, and load balancer all refer to the same thing. The nodes inside an LVS cluster are called real servers, and the puters that connect to the cluster to request its services are called client puters. The client puters, the Director, and the real servers municate with each other using IP addresses the same way puters have always exchanged packets over a work。 however, to make it easier to discuss this work munication, the LVS munity has developed a naming convention to describe each type of IP address based on its role in the work conversation. So before we consider the different types of LVS clusters and the choices you have for distributing your workload across the cluster nodes (called scheduling methods), let39。s look at this naming convention and see how it helps describe the LVS cluster. LVS IP Address Name Conventions In an LVS cluster, we cannot refer to work addresses as simply IP addresses. Instead, we must distinguish between different types of IP addresses based on the roles of the nodes inside the cluster. Here are four basic types of IP addresses used in a cluster: Virtual IP (VIP) address The IP address the Director uses to offer services to client puters Real IP (RIP) address The IP address used on the cluster nodes Director39。s IP (DIP) address The IP address the Director uses to connect to the D/RIP work Client puter39。s IP (CIP) address The IP address assigned to a client puter that it uses as a source IP address for requests sent to the cluster The Virtual IP (VIP) The IP address that client puters use to connect to the services offered by the cluster are called virtual IP addresses (VIPs). VIPs are IP aliases or secondary IP addresses on the NIC that connects the Director to the normal, public work.[1] The LVS VIP is important because it is the address that client puters will use when they connect to the cluster. Client puters send packets from their IP address to the VIP address to access cluster services. You tell the client puters the VIP address using a naming service (such as DNS, DDNS, WINS, LDAP, or NIS), and this is the only name or address that client puters ever need to know in order to use the services inside the cluster. (The remaining IP addresses inside the cluster are not known to the client puter.) A single Director can have multiple VIPs offering different services to client puters, and the VIPs can be public IP addresses that can be routed on the Inter, though this is not required. What is required, however, is that the client puters be able to access the VIP or VIPs of the cluster. (As we39。ll see later, an LVSNAT cluster can use a private intra IP address for the nodes inside the cluster, even though the VIP on the Director is a public Inter IP address.) The Real IP (RIP) In LVS terms, a node offering services to the outside world is called a real server. (We will use the terms cluster node and real server interchangeably throughout this book.) The IP address used on the real server is therefore called a real IP address (RIP). The RIP address is the IP address that is permanently assigned to the NIC that connects the real server to the same work as the Director. We39。ll call this work cluster work or the Director/realserver work (D/RIP work). The Director uses the RIP address for normal work munication with the real servers on the D/RIP work, but only the Director needs to know how to talk to this IP address. The Director39。s IP (DIP) The Director39。s IP (DIP) address is used on the NIC that connects the Director to the D/RIP work. As requests for cluster services are received on the Director39。s VIP, they are forwarded out the DIP to reach a cluster node. As is discussed in Chapter 15, the DIP and the VIP can be on the same NIC. The Client Computer39。s IP (CIP) The client puter39。s IP (CIP) address may be a local, private IP address on the same work as the VIP, or it may be a public IP address on the Inter. Types of LVS Clusters Now that we39。ve looked at some of the IP address name conventions used to describe LVS clusters, let39。s examine the LVS packetforwarding methods. LVS clusters are usually described by the type of forwarding method the LVS Director uses to relay ining requests to the nodes inside the cluster. Three methods are currently available: Network address translation (LVSNAT) Direct routing (LVSDR) IP tunneling (LVSTUN) Although more than one forwarding method can be used on a single Director (the forwarding method can be chosen on a pernode basis), I39。ll simplify this discussion and describe LVS clusters as if the Director is only capable of using one forwarding method at a time. The best forwarding method to use with a Linux Enterprise Cluster is LVSDR (and the reasons for this will be explained shortly), but an LVSNAT cluster is the easiest to build. If