Any instance in a physical cluster can be a failover resource. Logical cluster failover rules do not impact infrastructure, lock remastering, or recovery. You can configure logical cluster failover by specifying:
Failover resources – specific instances or groups of instances to which failover occurs.
The failover mode – determines whether failover occurs for individual members of a logical cluster or only for the cluster as a whole. See “Failover mode”.
The fail_to_any attribute – determines whether failover occurs only to designated failover resources or to any other instance if the designated failover resources are unavailable. See “Fail_to_any attribute”.
When you create a logical cluster, default settings for the fail_to_any attribute ensure that if a base instance fails, it is immediately replaced with another instance. This is the simplest failover strategy, and it is adequate for many sites.
If your site requires finer control of failover resources, you can change the default settings and direct failover to specific instances or groups of instances.
You can create up to 31 failover groups for each logical cluster. By grouping failover instances, you can give preference to specific failover groups. For example, you can ensure that instances in group 1 are considered before instances in group 2, and so on. The workload manager chooses failover instances within the group according to workload: instances with the lighter load are chosen first.
Instances can be a member of only one failover group per logical cluster. Thus, if instance “ase4” is in failover group 1 for “SalesLC”, it cannot also be in failover group 2 for “SalesLC”. However, “ase4” can simultaneously be in failover group 1 for “SalesLC”, failover group 2 for “HRLC”, and a base instance for “CatchallLC”.
When the workload manager needs to activate failover instances, it looks first in group 1, then in group 2, and so on until the failover condition is satisfied. If it cannot activate a configured failover resource, the workload manager checks the fail_to_any parameter configuration. If fail_to_any is true, the workload manager attempts to satisfy failover using any available instance. If fail_to_any is false, failover does not occur.
