Optimize Distributed Workload - Expert Oracle RAC Performance Diagnostics and Tuning - page 495

Table 15-1. ( continued )

Event Identifier Description

CARDINALITY It represents the service membership cardinality. It is the number of members that are running

the service. It can be used by client applications to perform software-based load balancing.

INCARNATION

Every time a member or node goes down and gets restarted, it gets a new cluster incarnation.

TIMESTAMP Server side date and time when the event was detected.

TIMEZONE The time zone given in GMT plus /minus value of Oracle Clusterware where the event occurred.

The example following is an event structure when an instance is brought down by a user operation:

INSTANCE VERSION=1.0 service=scdb database=scdb instance=SCDB_2 host=ssky1l1p1 status=down
reason=USER timestamp=2014-02-04 00:39:07 timezone=-05:00 db_domain= reported=date

When the preceding event is received by an ONS client machine, the application will use this information to
reroute any future connections to this instance until the load profile defined for the service has been met.
Oracle defines services for all components within the RAC environment to monitor its state and to notify the
application or client nodes of state changes. Figure 15-4 illustrates the relationship between the various components
that affect the application servers directly and are all monitored by ONS for state change notifications.

Server Pools INSTANCE

NODE

DATABASE SERVICEMEMBER SERVICE

RELOCATE

Figure 15-4. ERD (entity relationship diagram) and dependency relationship of application services

The number of components or subcomponents affected by a DOWN event depends on the component or service
that has failed. For example, if a node fails or is removed from the cluster membership, then the NODE event is sent
to all clients registered with the ONS of the failed node. All components that have a direct or indirect dependency to
the NODE are all affected. For instance, illustrated in Figure 15-4 , all entities database, instance, and the services that
the instance supports are all affected. Whereas a node or server pool cannot failover in the cluster, certain services,
database instance, including the database, can failover or relocate to another node in the cluster. 1 The type of service
defined as SINGLETON can failover or relocate when a node fails. Similarly, if the database is down or has relocated
itself to another node, then all services that depend on the database service will fail, and ONS will send that many
notifications to all participating nodes.

1 This requires that the database be configured as policy managed and not admin managed.