Secondary Database Availability

Replay of a schema instantiation on a secondary database has an online and an offline phase. When an online reorganization is replayed on an SDS secondary, the database remains available to inquiry applications for the duration of the online phase of the schema instantiation. Inquiry applications can continue to access objects defined by the current schema version.

When the initiate transition audit record is replayed, the database tracker performs the offline phase of the instantiation, accomplishing a transition to the latest version of the schema definition on the primary database. Initiating a transition has the same prerequisites as it did on the primary; that is, no versioned schemas and classes can be in use. To achieve this, JADE attempts to gracefully shut down all user applications, including server applications, before proceeding.

As the instantiation does not proceed until all user applications have been shut down, if a user application has stalled and is unable to terminate, the instantiation cannot proceed. To resolve a stalled instantiation, an administrative user must intervene to arrange for applications to be terminated and nodes to be shut down, if required. If necessary, you can shut down and restart the database server at this point.

When the schema instantiation has been replayed and committed, database server and application server applications configured to start automatically are restarted.

The following image shows the phases occurring on the secondary database.

By default, the initiate transition audit record is replayed as soon as it is encountered. If you need to schedule when this occurs so that you can arrange for a graceful termination of user applications, consider initiating a managed transition. The main difference in a managed transition is that a stop tracking audit record is inserted in the audit stream. This causes secondary database tracking to halt just prior to starting the offline phase so that the required downtime for inquiry applications can be managed.

The following image shows the phases in a managed transition.

Deferring replay by disabling tracking allows a failed schema instantiation to be spliced out on both the primary and secondary without the need to re-establish the secondary from a backup.

For details about starting applications in a secondary system and enabling user sign-on, see "Stopping and Starting Server Applications in Secondary Systems", in Chapter 10.