8 Using Oracle RAC Features

Overview of Fast Connection Failover

The Fast Connection Failover (FCF) feature is a Fast Application Notification (FAN) client implemented through the connection pool. The feature requires the use of an Oracle JDBC driver and an Oracle RAC database or an Oracle Restart on a single instance database. This section only describes the steps that an application must perform when using FCF with Oracle RAC. For more information on setting up an Oracle RAC database, see the Oracle Real Application Clusters Administration and Deployment Guide or consult an Oracle database administrator.

FCF manages pooled connections for high availability and provides the following features:

• FCF supports unplanned outages. Dead connections are rapidly detected and then the connections are aborted and removed from the pool. Connection removal relies on abort to rapidly sever socket connections in order to prevent hangs. Borrowed and in-use connections are interrupted only for unplanned outages.

• FCF supports planned outages. Borrowed or in-use connections are not interrupted and closed until work is done and control of the connection is returned to the pool.

• FCF encapsulates fatal connection errors and exceptions into the isValid API for robust and efficient retries. See "Checking If a Connection Is Valid" for more information on using this API.

• FCF recognizes new nodes that join an Oracle RAC cluster and places new connections on that node appropriately in order to deliver maximum quality of service to applications at run time. This facilitates middle-tier integration of Oracle RAC node joins and work-request routing from the application tier.

• FCF distributes run-time work requests to all active Oracle RAC instances.

Unplanned Shutdown Scenarios

FCF supports unplanned shutdown scenarios by detecting and removing stale connections to an Oracle RAC cluster. Stale connections include connections that do not have a service available on any instance in an Oracle RAC cluster due to service-down and node-down events. Borrowed connections and available connections that are stale are detected, and their network connection is severed before removing them from the pool. These removed connections are not replaced by the pool. Instead, the application must retry connections before performing any work with a connection.

Planned Shutdown Scenarios

FCF supports planned shutdown scenarios where an Oracle RAC service can be gracefully shutdown. In such scenarios, stale borrowed connections are marked and will only be aborted and removed after they are returned to the pool. Any on-going transactions do not see any difference and proceed to complete.

The primary difference between unplanned and planned shutdown scenarios is how borrowed connections are handled. Stale connections that are idle in the pool (not borrowed) are removed in the same manner as the unplanned shutdown scenario.

Starting from Oracle Database 12c Release 1 (12.1.0.2), UCP for JDBC supports graceful connection draining from any planned-down Oracle RAC instance. Affected borrowed connections are removed smoothly over a grace period, instead of immediate removal upon their return to the pool. This helps in avoiding throughput impact and logon storms during any service relocation.

Oracle RAC Instance Rejoin and New Instance Scenarios

FCF supports scenarios where an Oracle RAC cluster adds instances that provide a service of interest. The instance may be new to the cluster or may have been restarted after a down event. In both cases, UCP for JDBC recognizes the new instance and creates connections to the node as required.

Understanding Fast Connection Failover

After Fast Connection Failover is enabled, the mechanism is automatic; no application intervention is needed. This section discusses how a connection failover is presented to an application and what steps the application takes to recover, in the following sections:

• What the Application Sees

• How It Works

Fast Connection Failover Prerequisites

Fast Connection Failover is available under the following circumstances:

• The Universal Connection Pool is enabled.

  Fast Connection Failover works in conjunction with the JDBC connection caching mechanism. This helps applications manage connections to ensure high availability.

• The application uses service names to connect to the database.

  The application cannot use service identifiers.

• The underlying database has Oracle Database 12c Release 1 (12.1) Real Application Clusters (Oracle RAC) capability or Oracle Data Guard configured with either single instance Databases or Oracle RAC.

  If failover events are not propagated, then connection failover cannot occur.

• Oracle Notification Service (ONS) is configured and available on the node where JDBC is running.

  JDBC depends on ONS to propagate database events and notify JDBC of them.

• The Java Virtual Machine (JVM) in which your JDBC instance is running must have oracle.ons.oraclehome set to point to your ORACLE_HOME.

Example Fast Connection Failover Configuration

The following example demonstrates a connection pool that uses the FCF feature. FCF is configured through a pool-enabled data source. The example includes enabling FCF, configuring the Oracle Notification Service (ONS) and configuring a connection URL. These topics are discussed after the example.

PoolDataSource  pds = PoolDataSourceFactory.getPoolDataSource();

pds.setConnectionPoolName("FCFSamplePool");
pds.setFastConnectionFailoverEnabled(true);
pds.setONSConfiguration("nodes=racnode1:4200,racnode2:4200\nwalletfile=
/oracle11/onswalletfile");
pds.setConnectionFactoryClassName("oracle.jdbc.pool.OracleDataSource");pds.setURL("jdbc:oracle:thin@(DESCRIPTION= "+
   "(LOAD_BALANCE=on)"+
   "(ADDRESS=(PROTOCOL=TCP)(HOST=racnode1) (PORT=1521))"+
   "(ADDRESS=(PROTOCOL=TCP)(HOST=racnode2) (PORT=1521))"+
   "(CONNECT_DATA=(SERVICE_NAME=service_name)))");
...

The isValid method of the oracle.ucp.jdbc.ValidConnection interface is typically used in conjunction with the FCF feature and is used to check if a borrowed connection is still usable after an SQL exception has been thrown due to an Oracle RAC down event. For example:

try {  conn = pds.getConnection;  ...}catch (SQLException sqlexc)
{
   if (conn == null || !((ValidConnection) conn).isValid())
        
   // take the appropriate action
   
...
conn.close();
}

For more information on the ValidConnection interface, see "Checking If a Connection Is Valid".

Enabling Fast Connection Failover

The FCF pool property is used to enable and disable FCF. FCF is disabled by default. The following example demonstrates enabling FCF as shown in Example 8-1.

pds.setFastConnectionFailoverEnabled(true);

Querying Fast Connection Failover Status

An application determines if Fast Connection Failover is enabled by calling OracleDataSource.getFastConnectionFailoverEnabled, which returns true if failover is enabled, false otherwise.

Configuring ONS

FCF relies on the Oracle Notification Service (ONS) to propagate database events between the connection pool and the Oracle RAC database. At run time, the connection pool must be able to setup an ONS environment. ONS (ons.jar) is included as part of the Oracle Client software. ONS can be configured using either remote configuration or client-side ONS daemon configuration. Remote configuration is the preferred configuration for standalone client applications. This section discusses the following topics:

• Overview of ONS Configuration File

• Remote Configuration

• Client-Side Daemon Configuration

<component>[<subcomponent>,...];<component>[<subcomponent>,...];...

[all,!topology]

logcomp=ons[all,!secure]

# This is an example ons.config file
#
# The first three values are required
localport=4100
remoteport=4200
nodes=racnode1.example.com:4200,racnode2.example.com:4200

Configuring the Connection URL

The connection URL of a connection factory must use the service name syntax when using FCF. The service name is used to map the connection pool to the service. In addition, the factory class must be an Oracle factory class. The following example demonstrates configuring the connection URL as shown in Example 8-1:

...
pds.setConnectionFactoryClassName("oracle.jdbc.pool.OracleDataSource");
pds.setURL("jdbc:oracle:thin@//host:port/service_name");
...

The following examples demonstrate valid connection URL syntax when connecting to an Oracle RAC database. Examples for both the Oracle JDBC thin and Oracle OCI driver are included. Notice that the URL can be used to explicitly enable load balancing among Oracle RAC nodes:

Valid Connection URL Usage

pds.setURL("jdbc:oracle:thin@//host:port/service_name"); 

pds.setURL("jdbc:oracle:thin@//cluster-alias:port/service_name"); 

pds.setURL("jdbc:oracle:thin:@(DESCRIPTION= "+
   "(LOAD_BALANCE=on)"+
   "(ADDRESS=(PROTOCOL=TCP)(HOST=host1)(PORT=1521))"+
   "(ADDRESS=(PROTOCOL=TCP)(HOST=host2)(PORT=1521))"+
   "(CONNECT_DATA=(SERVICE_NAME=service_name)))");

pds.setURL("jdbc:oracle:thin:@(DESCRIPTION= "+
   "(ADDRESS=(PROTOCOL=TCP)(HOST=cluster_alias) (PORT=1521)) "+
   "(CONNECT_DATA=(SERVICE_NAME=service_name)))");

pds.setURL("jdbc:oracle:oci:@TNS_ALIAS");

pds.setURL("jdbc:oracle:oci:@(DESCRIPTION= "+
   "(LOAD_BALANCE=on) "+
   "(ADDRESS=(PROTOCOL=TCP)(HOST=host1) (PORT=1521)) "+
   "(ADDRESS=(PROTOCOL=TCP)(HOST=host2)(PORT=1521)) "+
   "(CONNECT_DATA=(SERVICE_NAME=service_name)))"); 

pds.setURL("jdbc:oracle:oci:@(DESCRIPTION= "+
   "(ADDRESS=(PROTOCOL=TCP)(HOST=cluster_alias) (PORT=1521)) "+
   "(CONNECT_DATA=(SERVICE_NAME=service_name)))");

Overview of Run-Time Connection Load Balancing

In an Oracle Real Application Clusters environment, a connection could belong to any instance that provides the relevant service. In the best case, all instances perform equally well and randomly retrieving a connection from the cache is appropriate. However, when one instance performs better than others, random selection of a connection is inefficient. The run-time connection load balancing feature enables routing of work requests to an instance that offers the best performance, minimizing the need to relocate work.

UCP JDBC connection pools leverage the load balancing functionality provided by an Oracle RAC database. Run-time connection load balancing requires the use of an Oracle JDBC driver and an Oracle RAC database. For more information on setting up an Oracle RAC database, see the Oracle Real Application Clusters Administration and Deployment Guide or consult an Oracle database administrator.

Run-time connection load balancing is useful when:

• Traditional balancing of workload is not optimal

• Requests must be routed to make optimal use of resources in a clustered database

• Capacity within the cluster differs and is expected to change over time

• The need to avoid sending work to slow, hung, and dead nodes is required

UCP for JDBC uses the Oracle RAC Load Balancing Advisory. The advisory is used to balance work across Oracle RAC instances and is used to determine which instances offer the best performance. Applications transparently receive connections from instances that offer the best performance. Connection requests are quickly diverted from instances that have slowed, are not responding, or that have failed.

Run-time connection load balancing provides the following benefits:

• Manages pooled connections for high performance and scalability

• Receives continuous recommendations on the percentage of work to route to database instances

• Adjusts distribution of work based on different back-end node capacities such as CPU capacity or response time

• Reacts quickly to changes in cluster reconfiguration, application workload, overworked nodes, or hangs

• Receives metrics from the Oracle RAC Load Balance Advisory. Connections to well performing instances are used most often. New and unused connections to under-performing instances will gravitate away over time. When distribution metrics are not received, connection are selected using a random choice.

Setting Up Run-Time Connection Load Balancing

Run-time connection load balancing requires that FCF is enabled and configured properly. See "Using Fast Connection Failover" for detailed instructions on setting up FCF.

In addition, you must configure the Oracle RAC Load Balancing Advisory with service-level goals for each service for which load balancing is enabled:

• The service goal must be set to one of the following:

  • DBMS_SERVICE.SERVICE_TIME

  • DBMS_SERVICE.THROUGHPUT

  The service goal can be set using the goal parameter, and the connection balancing goal can be set using the clb_goal parameter.

• The connection balancing goal must be set to SHORT. For example,

  EXECUTE DBMS_SERVICE.MODIFY_SERVICE (service_name => 'sjob' -, goal => 
     DBMS_SERVICE.GOAL_THROUGHPUT -, clb_goal => DBMS_SERVICE.CLB_GOAL_SHORT);
  

  Or

  EXECUTE DBMS_SERVICE.MODIFY_SERVICE (service_name => 'sjob' -, goal => 
     DBMS_SERVICE.GOAL_SERVICE_TIME -, clb_goal => DBMS_SERVICE.CLB_GOAL_SHORT);
  

The connection balancing goal can also be set by calling the DBMS_SERVICE.create_service procedure.

Overview of Connection Affinity

UCP JDBC connection pools leverage affinity functionality provided by an Oracle RAC database. Connection affinity requires the use of an Oracle JDBC driver and an Oracle RAC database version 11.1.0.6 or higher.

Connection affinity is a performance feature that enables a connection pool to select connections that are directed at a specific Oracle RAC instance. The pool uses run-time connection load balancing (if configured) to select an Oracle RAC instance to create the first connection and then subsequent connections are created with an affinity to the same instance.

UCP JDBC connection pools support the following two types of connection affinity:

• Transaction-Based Affinity

• Web Session Affinity

Transaction-Based Affinity

Transaction-based affinity is an affinity to an Oracle RAC instance that can be released by either the client application or a failure event. Applications typically use this type of affinity when long-lived affinity to an Oracle RAC instance is desired or when the cost (in terms of performance) of being redirected to a new Oracle RAC instance is high. Distributed transactions are a good example of transaction-based affinity. XA connections that are enlisted in a distributed transaction keep an affinity to the Oracle RAC instance for the duration of the transaction. In this case, an application would incur a significant performance cost if a connection is redirect to a different Oracle RAC instance during the distributed transaction.

Web Session Affinity

Web session affinity is an affinity to an Oracle RAC instance that can be released by either the instance, a client application, or a failure event. The Oracle RAC instance uses a hint to communicate to a connection pool whether affinity has been enabled or disabled on the instance. An Oracle RAC instance may disable affinity based on many factors, such as performance or load. If an Oracle RAC instance can no longer support affinity, the connections in the pool are refreshed to use a new instance and affinity is established once again.

Applications typically use this type of affinity when short-lived affinity to an Oracle RAC instance is expected or if the cost (in terms of performance) of being redirected to a new Oracle RAC instance is minimal. For example, a mail client session might use Web session affinity to an Oracle RAC instance to increase performance and is relatively unaffected if a connection is redirected to a different instance.

Setting Up Connection Affinity

Perform the following steps to set up connection affinity:

• Enable FCF.

  See Also:

  "Using Fast Connection Failover"

• Enable run-time connection load balancing.

  See Also:

  "Using Run-Time Connection Load Balancing"

• Create a connection affinity callback.

• Register the callback.

This section contains the following subsections:

• Creating a Connection Affinity Callback

• Registering a Connection Affinity Callback

• Removing a Connection Affinity Callback

public class AffinityCallbackSample
   implements ConnectionAffinityCallback {
   
   Object appAffinityContext = null;
   ConnectionAffinityCallback.AffinityPolicy affinityPolicy =
   ConnectionAffinityCallback.AffinityPolicy.TRANSACTION_BASED_AFFINITY;
   
   //For Web session affinity, use WEBSESSION_BASED_AFFINITY;
   
   public void setAffinityPolicy(AffinityPolicy policy)
   {
      affinityPolicy = policy;
   }
   
   public AffinityPolicy getAffinityPolicy()
   {
      return affinityPolicy;
   }
   
   public boolean setConnectionAffinityContext(Object affCxt)
   {
      synchronized (lockObj)
      {
         appAffinityContext = affCxt;
      }
      return true;
   }
   
   public Object getConnectionAffinityContext()
   {
      synchronized (lockObj)
      {
         return appAffinityContext;
      }
   }
}

ConnectionAffinityCallback callback = new MyCallback();

PoolDataSource  pds = PoolDataSourceFactory.getPoolDataSource();

pds.setConnectionPoolName("AffinitySamplePool");
pds.registerConnectionAffinityCallback(callback);
...

PoolDataSource  pds = PoolDataSourceFactory.getPoolDataSource();

pds.setConnectionPoolName("AffinitySamplePool");
pds.removeConnectionAffinityCallback();
...

Overview of Global Data Services

Global Data Services (GDS) is a new feature introduced in Oracle Database 12c Release 1 (12.1). Through this feature, Fast Connection Failover, Run-time Connection Load-Balancing, and Connection Affinity features that were available only in Oracle RAC, are now extended to a set of replicated databases that offer common services.

The set of databases may include Oracle RAC and single-instance Oracle databases interconnected through Data Guard, GoldenGate, or any other replication technology. A database service that can be provided by multiple databases is called a global service, so that it can be distinguished from the traditional service that can be provided only by a single database. This combination enables services to be deployed anywhere within this globally distributed configuration, supporting load balancing, high availability, database affinity, and so on.

Configuration Settings of Applications Using GDS

UCP connects to Global Data Services in the same way that it connects to local services on an Oracle RAC. The service name in the connection string should be the name of the global service. The endpoint should be the endpoint of a GDS listener instead of the endpoint for the local, remote, or SCAN listener of a database.

A client must specify its region in the REGION parameter of the connection string. This is a new requirement for GDS. The region name is required because, in case of GDS, Run-time Load Balancing advisory is customized for particular regions. Following is an example of a typical connection string:

(DESCRIPTION=
  (ADDRESS=(GDS_protocol_address_information))
  (CONNECT_DATA=
   (SERVICE_NAME=global_service_name)
   (REGION=region_name)))

Like with local services, UCP can specify multiple GDS listeners in the same connection string for listener failover, load balancing, or both.

(DESCRIPTION=
  (ADDRESS_LIST=
    (LOAD_BALANCE=ON)
    (FAILOVER=ON)
    (ADDRESS=(GDS_protocol_address_information))
    (ADDRESS=(GDS_protocol_address_information)))
  (CONNECT_DATA=
   (SERVICE_NAME=global_service_name)
   (REGION=region_name)))

The REGION parameter is optional if only global service managers from the local region are specified in the client connection string. This is the case when there is only one region in the GDS configuration, or can be the case when there are multiple regions. But, it is not feasible to change the connection string of the an existing client designed to work with a single database. If the REGION parameter is not specified, then the client's region is assumed to be the region of the global service manager used to connect to the global service.

All GDS listeners in the preceding example belong to the same region where UCP is running, that is the local region. To provide high availability, when all GDSs in the local region are unavailable, you can specify the GDS listeners for the buddy region in additional ADDRESS_LIST descriptors.

(DESCRIPTION=
  (FAILOVER=on)
  (ADDRESS_LIST=
    (LOAD_BALANCE=ON)
    (ADDRESS=(global_protocol_address_information))
    (ADDRESS=(global_protocol_address_information)))
  (ADDRESS_LIST=
    (LOAD_BALANCE=ON)
    (ADDRESS=(global_protocol_address_information))
    (ADDRESS=(global_protocol_address_information)))
  (CONNECT_DATA=
   (SERVICE_NAME=global_service_name)
   (REGION=region_name)))

You do not need manual ONS configuration because UCP automatically retrieves the ONS connection information that is optimally customized for the UCP region from GDS.