Design And Implementation Of Linux Based High-Availability Cluster Management System

The assurance of service continuity is becoming more and more important withthe extensive application of computer networks．The server systems, for example, infields of finance, telecommunication and government, are heavily burdened with thetasks of information storing, processing, and delivering, etc. Once the server systemsare out of order, huge losses may be caused. The high-availability clusteringtechnology must be used to guarantee the service continuity. High-availability clustercan be employed to provide uninterrupted services, and to achieve fault detection andautomatic task switch, which reduces the loss caused by computer hardware andsoftware faults, or human errors to minimum．Therefore; high-availability clusteringtechnology is one of the hot research subjects.In this thesis, we first summarize and analyze the state of art and theimplementation technologies of the existing high-availability cluster system. Then weproposes a scheme of building high-availability cluster by using open-source code, andgives the design and implementation of a high-availability cluster system based onLinux. The system achieves high-availability by adding a backup node for the primarynode and adopting the popular dual-machine backup technology. This system employsthe open-source database of PostgreSQL, which exploits dual-machine backupmechanism. Therefore, it simplifies the cluster management and releasing the clustersystem from the problems of data synchronization between the master and slave node.It uses Heartbeat to detect the node states and determine the active node between theprimary node and backup node in the cluster. If the primary node fails, in order toachieve uninterrupted services provide by the cluster, the backup node will take overthe task of cluster management, and start the primary service by using Pacemaker. Thecluster management system has the characteristics of convenient management,high-availability and low deployment cost. The testing results show that the system canswitch between the nodes and takes over services efficiently, and achieveshigh-availability of the cluster.