Linux cluster programming for high availability with an Oracle instance

Abstract

This paper presents programming performed in a cluster of two Linux servers to ensure the high availability of an Oracle instance. The problem of database downtime is proposed to be solved when a component of a server or the entire server fails and there is no service to the database users. The objective was to develop programming to start, stop, and monitor the status of an Oracle instance on one of the two Linux servers in a cluster in case of contingency or maintenance of the other server. The methodology followed divided the programming into five modules: cluster manager, package manager, monitoring module, Host Bus Adapter (HBA) port manager, and network port manager. High availability was achieved by creating a package that contained the Oracle instance and resources to start it on the cluster servers. The package is assigned an IP address to which users of the instance are connected. The contribution of this work is to provide a low-cost solution, compared to existing commercially similar systems, with a quick response and easy implementation that allows a company or institution to continue working after a hardware or software failure. The startup time of the Oracle instance package after contingency on a server was 20 s, which was the time when the cluster application was not available to the user. The results of the methodology were a cluster that eliminates the points of failure represented by LAN ports, HBAs, hard disk drives and an entire server on which a mission-critical application is running. Without the use of the cluster, the user must wait for the repair of the failed component(s) to restore the company's operation. The cost of the cluster is 10% of the cost of an equivalent commercially available solution. The application failover time is 20 s, which is one-tenth of the time achieved in other solutions that use a proprietary operating system.