To configure the Cluster Edition for a round-robin method, use:
sp_addserver 'SYB_BACKUP', NULL, '$ROUNDROBIN'
Once you configure the Cluster Edition to use the round-robin method, the Cluster Edition selects a Backup Server for a dump or load command according to which Backup Server that is performing the least amount of work.
For example, if a user connects to instance Inst1 from isql and issues a dump command from this session, instance Inst1 checks the SYB_BACKUP entry in sysservers, and finds that it points to keyword $ROUNDROBIN. If Inst1_BS is busy performing another task, the Cluster Edition moves to the next Backup Server, Inst2_BS. If Inst2_BS is available, with no jobs running on it, this Backup Server performs the dump, and its index is incremented to indicate that it is busy and unavailable for work. However, if Inst2_BS is busy, the Cluster Edition moves to the next Backup Server, and so on until it finds a free Backup Server.
On the other hand, if the connection attempt to Inst2_BS fails, then the enable backupserver HA parameter governs whether or not the Backup Server is restarted automatically or the next available Backup Server is attempted. An error is reported only if all configured Backup Servers are unavailable.
If you set enable backupserver HA to true and configure Backup Server for a round robin policy, you must enable all nodes to start any Backup Server in the cluster. This is necessary to allow another node to perform the dump or load duties if the initial node fails on which you issued the command. To enable all nodes to start any Backup Server, include entries for all Backup Servers in the interfaces file.
For example, Big_Cluster uses a round-robin policy, and includes instances Node_1, Node_2, and Node_3. The interfaces file entry for instance Node_1 must include entries for Backup Servers running on nodes Node_2 and Node_3 in case the round robin policy select either of these as the replacement node if Node_1fails:
Node_1_BS master tcp ether Node_1 5004 query tcp ether Node_1 5004 master tcp ether Node_2 5006 query tcp ether Node_2 5006 master tcp ether Node_3 5008 query tcp ether Node_3 5008 Node_2_BS master tcp ether Node_1 5004 query tcp ether Node_1 5004 master tcp ether Node_2 5006 query tcp ether Node_2 5006 master tcp ether Node_3 5008 query tcp ether Node_3 5008 Node_3_BS master tcp ether Node_1 5004 query tcp ether Node_1 5004 master tcp ether Node_2 5006 query tcp ether Node_2 5006 master tcp ether Node_3 5008 query tcp ether Node_3 5008
Alternatively, you can start Backup Servers manually on the node on which you are issuing the dump or load command.
