Double-haired hot standby with Windows2000
Author: Tang Hua Transfer: Thousand blog www.2ok.cn the occasion in the traditional mode of network terminal / mainframe era, the terminal features a simple, maintenance work, specialized in the management and maintenance of end hosts, with resource sharing, manageable Features. However, the main unit cost is expensive, and the terminal has no handling ability, limiting the scale development of the network. The subsequent client / server model has advanced the development of standardization and development of the computing industry, providing a considerable flexibility for the system, but with the scale of the distribution system, the system maintenance and management has brought huge overhead. . In the server cluster system, the server is no longer distributed throughout, but is centrally managed and maintained. It maintains the development, scalability of distributed client / server mode, while also has the advantages of resource sharing and concentration of terminal / host mode. In a relatively concentrated cluster system, it reduces the cost of system management, and also provides processing capabilities comparable to large server systems. The service-oriented application for Internet needs high-performance hardware platforms as support, and the parallel technology is applied in the server field, which is an inevitable trend of computer development. Parallel processing technology In high performance calculations, high availability and high performance are two important directions of cluster server system development. The concept of clustering clusters The English name is Cluster, which is independent of each other. It is configured by a high-speed network interconnected computer, which is configured with a single system mode. When a customer interacts with the cluster, the cluster is a separate server. Cluster configuration is used to improve availability and scalability. Compared with traditional high-performance computer technology, clustering technology can utilize servers of each grade as nodes. The system cost is low. It can achieve high computing speed. Complete the calculation of the main running calculation, with high response ability, can meet today The demand for increasing information services. The number of cluster technology applications The number of Internet users has a geometric grade growth and the complexity of scientific calculations requires higher processing capabilities, while the development of CPUs cannot keep up with the growing demand, so we face the following questions: ● large scale calculation As gene data analysis, meteorological prediction, oil exploration requires extremely high computational performance. ● The development of application scale makes a single server to undertake load. ● Changging demand requires hardware with flexible scalability. ● Key business requires a reliable fault tolerance mechanism. Ia Cluster System (Cluster) Features ● Computed a unified computer system by a number of complete computers; Special software support, such as operating systems or databases that support cluster technology, etc., single-system images, ie operation control, IP login point, file structure, storage space, I / O space, job management system, etc. ● You can dynamically add new servers and delete servers that need to be eliminated in the cluster system to maximize the need to maximize the needs of growing applications; ● Availability is the most important factor in cluster system applications. And an important indicator of the system; ● Can provide uninterrupted services to users, because multiple nodes include multiple nodes in the system, the entire system can continue to provide services to users; ● Have pole High performance price ratio, compared with traditional large hosts, with great price advantage; ● Resource makes full use, each node of the cluster system is relatively independent machine, when these machines do not provide service or no need When using it, it is still fully utilized. The accessories that have been updated on large hosts are difficult to be reused.
Implement server cluster hardware configuration ● Two web servers ● Server operating system hard disk two pieces ● Ultra 160 LVD SCSI disk array One ● 18g SCSI hard disk three pieces ● Network service network card four server cluster Practice step ● Before installing the cluster service Prepare: 1, three 18g SCSI hard drives constitute a disk array, do RAID5. 2, two server requirements are configured with dual network cards, install the Microsoft Windows 2000 Advance Server operating system, and configure the network. 3. All disks must be set to basic disk, and the array disk partition must be greater than 2. 4. Each server is to join the domain, become a domain member, and have administrator privileges on each server. ● Install the configuration server network key 1, in this part, each server requires two network adapters, a connection public network, a connection internal network (it only contains the cluster node). Internal network adapter establishes a communication, cluster status signal, and cluster management of point-to-point points. The public network adapter of each node is connected to the cluster to the public network and resides here. 2. After installing the Microsoft Windows 2000 Adwan Server operating system, start configuring the network of each server. In the network connection, we named the "external network" to the public network, connect the internal network name "intranet" and specify the IP address as: Node 1: Introduction: IP: 10.10.10.11 Web IP: 192.168 .0.192 Subnet mask: 255.255.255.0 Gateway: 192.168.0.191 (Main domain controller IP) Node 2: Inner Web: IP: 10.10.10.12 Outer network IP: 192.168.0.193 Subnet mask: 255.255.255.0 Gateway: 192.168.0.191 ● Install the configuration array 1, and make the array to verify RAID5. And divided into partitioning. Create a small partition (at least 50 mega) to store the cluster configuration database detection point and log file. This is an important key partition in the cluster, one but he has failed, will lead to the failure of the entire cluster. Here we allocate 1G space and name X drive. 2, other partitions can be set according to the needs of users. We only divide a Y area for convenience. 3, Note: All disks in the cluster must be basic disk and have NTFS format. ● Install cluster service software [Configure the first patch] 1, all devices are first closed (two servers, one disk array). 2, first open the disk array (turn on the disk array first at all times to avoid disk confusion). 3. Open the node 1, log in to the domain with the user name with administrator privileges, add the component - cluster service with the Add Remove program. Select this is the first point of the cluster during the installation process. Create a cluster name. We named mycluster. 5, configure the cluster disk. Since we only have a shared disk array, we configure both partitions in this array to a group disk. 6, configure the cluster network. First we have the external network and the network name is selected. The IP address is held and the IP address of the external network is consistent. Select the network to be used for cluster services. Check all communication (mixing networks) options. Perform the next step. Next, configure the intranet to ensure that the network name and IP address are consistent with the intranet, and the network can be used for cluster services. Check only for internal cluster communication options. 7. Click the next step button, the screen appears in the order of the network. We adjust the first connection to the table is the intranet.
This is because: In a normal operation, the connection will be used for cluster communication. If the internal network connection is invalid, the cluster service will automatically switch to the next network in the table. 8. Perform the next step will appear the cluster IP and mask options. Here, enter the only cluster IP, we set to 192.168.0.195 mask of course: 255.255.255.0. 9. Click Finish, complete the cluster configuration of the first node. [Configure the second node] 1, keep the array and the first node boot and available. The second node is electrically opened. 2, like the cluster service is installed on the first node, the second node begins to install. 3, the process is the same in the first node, only the following differences: In the Create or Add Cluster dialog box, select the second or next node in the cluster, and click Next. Enter the cluster name created earlier (in this case, it is mycluster), and click Next. Set it to the cluster to unchecked. The Cluster Service Configuration Wizard will be automatically supplied and the user account name selected when installing the first node. Enter the password of the account (if any), and click Next. In the next dialog, click Finish and end the configuration. The cluster service will be started. Click OK. Close Add / Remove Programs If you want to install additional nodes, repeat these steps to install cluster services on other nodes. Check installation Click Start, click Programs, click Administrative Tool, and click Cluster Manager. The two nodes shown indicate that there is a cluster, and it is running. Effectiveness evaluation ● The test of fault tolerance is the following experiments for his fault tolerance. Node, at which time the cluster service running on this node is persecuted and transferred to another node connected to the normal connection, this process is about 4 seconds (fixed to the number of services running on the cluster, in installation SQL Server) This process is extended to 9-15 seconds after the 2000 service. 2, in the case of simultaneous operation of the two nodes and cluster services, we changed the time of a node, and automatically adjusts the same settings as the other node after approximately 10 minutes. 3. We open the web service on the cluster service and set up the download project. At the client is downloading a node in the cluster, the client does not have a significant change. It can be seen from the above experiments that the cluster made with Windows 2000 Advance Server has good performance in fault tolerance. He is almost instantly take over the service on the wrong node, and there is almost no change relative to the client. At the same time, he synchronized all the nodes in the cycle time to ensure its consistency. ● Load balancing capabilities 1, the cluster made by Windows 2000 Advance Server is based on TCP / IP construction, and he provides high availability and scalability for a WEB server that works together and two or more host clusters. Internet customers use a single IP address to access the cluster. Customers cannot distinguish between a single server from the cluster. Server programs cannot recognize they are running in a cluster. However, due to network load balancing clusters, uninterrupted services can be provided even if the cluster host has failed, it is a single host that runs a single server program. Compared with a single host, the cluster can also make a more rapid response to customer needs. Note that we have encountered some problems in the process of being a clustered experiment. Although they all succeeded smoothly, I think I still have to mention it here, let everyone pay attention to avoid unnecessary repetition errors in future work. 1. The importance of the X disk in the shared disk.
