Restrictions and Table Requirements for Classic Replication Schemes

All masters and subscribers must have their clocks synchronized through NTP or other means. The clock skew between all masters and subscribers cannot exceed 250 milliseconds. When adjusting the system clocks on any nodes to be synchronized with each other, do not set any clock backward in time.

The name and owner of replicated tables participating in the classic replication scheme must be identical on the master and subscriber databases. However, the definition for the columns of replicated tables participating in the replication scheme do not necessarily need to be identical. For more information on the column definition options, see Column Definition Options for Replicated Tables.

Replicated tables must have one of the following:

• A primary key

• A unique index over non-nullable columns

Replication uses the primary key or unique index to uniquely identify each row in the replicated table. Replication always selects the first usable index that turns up in a sequential check of the table's index array. If there is no primary key, replication selects the first unique index without NULL columns it encounters. The selected index on the replicated table in the master database must also exist on its counterpart table in the subscriber.

Note:

The keys on replicated tables are transmitted in each update record to the subscribers. Smaller keys are transmitted more efficiently.

Replicated tables have these restrictions:

• A primary key column cannot have a LOB data type.

• You cannot replicate tables with compressed columns.

If these requirements and restrictions present difficulties, you may want to consider using the Transaction Log API (XLA) as a replication mechanism. See Using XLA as a Replication Mechanism in Oracle TimesTen In-Memory Database C Developer's Guide.