The Applier uses recovery tables for all target types except

Apache Kafka,

Cloudera,

Flat File,

and Hortonworks

.

For Kafka targets, Data Replicaton uses a checkpoint file instead of a recovery table.

The Applier writes information to recovery tables every apply cycle. When an apply cycle starts, if the previous apply cycle completed successfully, the Applier purges the data in the recovery table. If the previous apply cycle ended abnormally, the Applier uses the data in the recovery table to recover from the failure.

The Applier uses data in the recovery table under the following circumstances:

During the recovery cycle, the Applier processes only the intermediate files that were not successfully processed during the previous apply cycle. If the recovery cycle completes successfully, the Applier continues to process the intermediate files.

Typically, the Applier generates recovery tables when it runs for the first time. To generate the recovery table on the target, the Applier must run under a user who has the privileges required to create a table on the target. When you save a configuration file for the first time or deploy a configuration to another Server Manager Main server, Data Replication displays a dialog box in which you can edit the recovery table name and confirm the generation of the recovery table. The default name of the recovery table is IDR_RECOVERY. For a replication job with multiple targets, you can override the recovery table name to specify a different name for each target.

For Netezza targets, Data Replication creates multiple recovery tables on each database. The number of recovery tables for a Netezza database must match the number of Applier threads. Data Replication appends the suffix _

N

to recovery table names on Netezza targets, where

N

is the sequence number of the Applier thread that uses this recovery table.

If an error occurs when the Applier is generating the recovery tables, you can manually define and execute CREATE TABLE statements to create recovery tables. For more information, see DDL Statements for Manually Creating Recovery Tables.

With both recovery tables and checkpointing in place, the Extractor, Server Manager, and Applier can resume processing correctly after different types of failures and planned outages, such as database failures, network failures, system failures, Data Replication upgrades, and configuration information upgrades.