2.19.4 Separating a node from a multi-node configuration

(1) When a node failure occurs on the master node

If a node failure occurs on the master node, the master node is separated from the multi-node configuration. Then, one of the slave nodes becomes the master node, and processing continues with the remaining nodes.

The following figure shows the process of separating a node if a node failure occurs on the master node.

Figure 2‒70: Process of separating a node if a node failure occurs on the master node

Explanation

• The master node on which a node failure occurred is separated from the multi-node configuration, and one of the slave nodes is switched over to the master node. This is called master node switchover.

• The SQL statements that were being executed on the master node on which a node failure occurred terminate in an error, and the application programs that were running transactions are disconnected from the HADB server. The application programs that were connected to the HADB server and were not running any transactions when the node failure occurred are also disconnected from the server.

Note

On the slave node that will become the new master node, the update processing that was running on the current master node, on which a failure occurred, is restored (rolled back). After restoration finishes, master node switchover takes place.

HA Monitor manages node separation and master node switchover (by using the hot standby function) in the event of a node failure. The master node operates as the active system of HA Monitor, and the slave nodes operate as the standby systems of HA Monitor.

When switching one of the slave nodes to the master node, HA Monitor also switches the shared disk, and adds and deletes the alias IP address.

Note

The slave node that the hot standby function of HA Monitor assigns as the new active system becomes the new master node. Which slave node becomes the new master node depends on the specifications of HA Monitor.

To Page Top

(2) When a node failure occurs on a slave node

If a node failure occurs on a slave node, the node is separated from the multi-node configuration, and processing continues with the remaining nodes.

The following figure shows the process of separating a slave node on which a node failure occurred.

Figure 2‒71: Process of separating a slave node on which a node failure occurred

Explanation

The SQL statements that were being executed on the slave node on which a node failure occurred terminate in an error, and the application programs that were performing transactions are disconnected from the HADB server.

Note

The transactions that were being executed on the failed slave node are rolled back on the master node.

To Page Top