To achieve any level of automatic recovery if a slave node fails, YARN must be used. By default, YARN is disabled. For details on how to enable YARN, see the VectorH User Guide.

Without YARN, if a slave node fails, the cluster must be manually stopped, re-configured to remove the failed node from the slaves file in $II_SYSTEM/ingres/files, and then re-started.

To simulate the failure of a slave node, the same reboot procedure can be used as for the master node.