Increased Server Availability and Flexibility through Failover Capability

As computing systems become increasingly mission-critical, a high level of service availability is essential. In order to maintain service availability, it is desirable to have the ability to migrate services from one server to another while having this change remain transparent to the client machines. Although commercial solutions exist which provide automatic failover capability, they are often costly and restrictive. Manual failover capability is useful for providing service availability in situations where there is a hardware failure, or where a server must be taken down for extended maintenance. The discussion in this document will explore what primitives can be used to construct a failovercapable system, the issues involved in any service migration, and some of the specific details about doing migration of services such as NFS, sendmail, and World Wide Web. Although implementation-specific examples provided assume a Sun Solaris 2.x operating environment, the use of a logical volume manager, and ethernet connectivity, other flavors of UNIX may also contain the necessary building blocks needed to build a failover-capable system.