For CDC sessions, you must select the recovery strategy of
Resume from the last checkpoint
in the session properties to use PowerCenter and PowerExchange restart and recovery functionality. With this recovery strategy, the PowerCenter Integration Service can recover target table data and PWXPC can recover the CDC restart information.
PowerExchange supplies CDC restart tokens that describe a specific start point in the change stream. For relational targets, PWXPC stores the restart tokens in the recovery state tables in the target databases. For nonrelational targets, PWXPC stores the restart tokens in the recovery state file on the PowerCenter Integration Service machine.
PWXPC issues the following message when the
Resume from last checkpoint
recovery strategy is enabled:
PWXPC_12094 [
process_id
] Advanced GMD recovery in effect. Recovery is automatic.
With this recovery strategy, after a failure, PWXPC can automatically recover sessions that are warm started.
If you use the
Resume from the last checkpoint
recovery strategy, the following considerations apply:
If you have a PowerCenter high availability (HA) system, you must enable HA recovery for your CDC workflows. Otherwise, when you warm start a CDC session, a restart error occurs and you must cold start the session. To configure HA recovery for a CDC workflow and avoid this error, in Workflow Manager, navigate to the
Properties
tab for the workflow and then select the
Enable HA Recovery
option.
If you use any of the following transformation types for your sessions, you must select the
Output is Deterministic
option for the transformation:
External Procedure transformation
HTTP transformation
Java transformation
Source Qualifier transformation
SQL transformation
Stored Procedure transformation
Unstructured Data transformation
If you use a Lookup transformation, select either the
Lookup Source is Static
or
Lookup Cache Persistent
option for the transformation instead.
If you do not select the
Output is Deterministic
option or the
Lookup Source is Static
or
Lookup Cache Persistent
option for the transformation when using the
Resume from last checkpoint
recovery strategy, you cannot validate the CDC session in Workflow Manager.
If you use the File Writer to write CDC data to flat files and other target types, such as relational targets, in the same session, do not enable resume-from-last-checkpoint recovery processing. If you do, the restart tokens might be compromised, causing possible data loss or duplication.
If you run a session with
Resume from last checkpoint
recovery strategy and the session fails, do not alter the mapping, session, or state information before you restart the session. PWXPC cannot guarantee recovery if you make any of these changes.
When the PowerCenter Integration Service resumes a session, it restores the session state of operation, including the state of each source, target, and transformation. PWXPC, in conjunction with the PowerCenter Integration Service, determines how much of the source data it needs to reprocess.