Transaction branch information is the information about back-end server names that are subject to replication. When the import transaction synchronization facility is used, the transaction branch information is stored in the source HiRDB's system log file. The source Datareplicator extracts that information and guarantees data integrity.

The following figure shows the linkage of transaction branch information.

Figure 3-46 Linkage of transaction branch information

#1

When a transaction is executed on the source HiRDB, transaction branch information is output to HiRDB's system log file for each transaction. At that point, the transaction branch information contains the names of all target back-end servers for the corresponding transaction.

#2

If transaction branch information has been output to the system log file, the source Datareplicator's extraction process unconditionally extracts the transaction branch information and stores it in the extraction information queue file. At this point, the names of the back-end servers that will update a table other than the tables to be extracted are deleted.

#3

The source Datareplicator's transmission process converts the back-end server names in the transaction branch information to the data linkage identifiers, and sends them to the target Datareplicator.

#4

Transaction branch information is not extracted for a back-end server that updates a table that is not the target of extraction.

HiRDB's normal termination log is replaced with a synchronous event and is then transmitted. This enables an import transaction to be synchronized even when normal termination of HiRDB is detected. Note that the synchronous event is extracted when HiRDB starts normally. The import transaction is not synchronized when HiRDB terminates normally.