Compaq ProLiant 6500 Deploying the ProLiant Cluster for NetWare 4.2 - Page 26

'HSOR\LQJ WKH 3UR/LDQW &OXVWHU IRU 1HW:DUH   Z.E.N.works Application Failover Any server applications accessing the shared drives after a failover must be started on the surviving server as described above, and any network drives will need to be re-mapped. It is possible using Novell Z.E.N.works with the Novell Application Launcher to set up an application in NDS with failover capabilities. Using Z.E.N.works, a network administrator can create an application object in NDS for each server and then allow the failover of the application in the NetWare Administrator to point to the other server's NDS object. At failover, the clients' application will disconnect or crash if it was using resources on the failed server, but the user will be able restart the application from the failed over server. The only data loss would be what was not saved before the application disconnects or crashes. IP only clients and applications will be disrupted during the failover process, but if the IP configuration information and application are transferred to the surviving server, the client will resume normal operation after failover. Disruptions During Failover With NHAS, the recovery time in the event of a failover depends on the types of applications running on the server. Client/server applications performing database transactions in which records are being accessed by the database, or word processing applications where the server is accessed only to read and write a document, may not see any problem during a server failure and recovery. Upon failover, some disruption to applications can occur. This may vary from a slight delay in accessing the recovery server to the need for complete re-logging into the system. Other applications, for example, database servers, must have clustering awareness built into them so that transactions can be rolled back and logs parsed to ensure data integrity is maintained. Oracle, on the other hand, features complex transaction logging systems that maintain the integrity of the database in the event of a failure, regardless of cluster awareness. But these interruptions are indeed trivial compared to the delay and cost of manually swapping in a new server and restoring the data and applications to the last backup. User Licenses in a Failover Novell High Availability Server's support for dynamic user licensing automatically adds the user license count from a failed server to your surviving server at failover time. When the failed server is brought back into operation, the user license count is automatically restored. Management Failover Once a server initiates the failover detection process, a failover can only be paused. During the failover countdown, you may press P to pause the countdown. However, once unpaused and having expired the failover delay, the server begins "Entering Failover." At this point, failover cannot be stopped. If Server A does not generate any traffic across any line of server-to-server communication because of an intensive thread or other condition, it is possible for Server B to detect that Server A has failed, if the period of traffic inactivity continues. This "partition in time" may actually initiate a failover and cause all volumes of Server A to be dismounted and then mounted as volumes on Server B. However, Server A has not truly failed. Moreover, since Server A has been too busy with this internal thread, it has not noticed that Server B has assumed control of its resources. 0096-0699-A