Compaq ProLiant 3000 Deploying the ProLiant Cluster for NetWare 4.2 - Page 25

'HSOR\LQJ WKH 3UR/LDQW &OXVWHU IRU 1HW:DUH   8. Initiate a failover to Server B in the NHAS console. In other words, we will cause Server A to fail and transfer all network services to Server B. For the purposes of this illustration, cause Server A to fail by powering off the server. Initiating a failover by pressing F at the console requires that the NHAS software be paused on the downed server to prevent immediate recovery. Powering the server off ensures that the failover will be successful and sustained. At this point, Server B will execute the FAIL2.NCF file which will assign Server A's former secondary IP address as a secondary IP address handled by Server B. Throughout the process, the client ping should be able to reach 123.45.67.11. After the nearly instantaneous transfer is complete, Server B alone will respond to any requests sent to both 123.45.67.11 and 123.45.67.22 In essence, the client is now talking to Server B exclusively. If a failback to Server A is initiated at the console, Server B executes the CO2.NCF script file, which deletes the 123.45.67.11 secondary IP transferred from Server A. At the same time, Server A will process the CO1.NCF file and, once again, handle requests for IP address 123.45.67.11. With the failback complete, Server B will handle only requests to 123.45.67.22. Notes on Failover As an emerging technology, clustering presents unique challenges to legacy applications and environments. While none of these challenges is prohibitive, certain steps or precautions are often necessary to accommodate the failover process. Server Applications and the SYS: Volume Some applications, such as Netscape Fast Track Server, write the executable files to the server's SYS: volume. In the event of a failover, the surviving server would no longer be able to access the files on the downed server's SYS: volume. If your ProLiant cluster runs applications that write to the server's SYS: volume, you must install the executables for that application on both servers. In addition, you must create a script that will launch any such applications on the surviving server after a failover. Client Drive Mappings with NDS and Novell Client 32 If Novell Client32 and NDS are being used, the clients will remain attached to the network during a cluster failover. Network drives that were mapped attached to drives previously owned by the failed server, though, will be broken. The existing Novell Client32 technology does not support automatic failover drive mappings. A failed-over shared drive can be re-mapped in Windows by using Windows Explorer as follows: 1. Right click on the Start button. 2. Select Explore. 3. Select Map Network Drive from the Tools menu. Then select a drive letter and specify the path to the data to be mounted as a drive on the client. 4. Finally, choose whether you want to reconnect this drive mapping. Then select OK. If the client applications are IP only, and the virtual IP addresses are transferred at failover to the surviving server, then these applications will reconnect after the failover is complete. 0096-0699-A