The Teradata Access Module for Named Pipes supports normal checkpoint and restart/recovery operations on the reader process system, but has no such interaction with the writer process.
- If the reader process terminates unexpectedly, restart the job, which causes the Teradata Access Module for Named Pipes to use the fallback data file to relocate to the last checkpoint in the data stream.
- If the writer process terminates unexpectedly, manually abort the reader process and resynchronize the job.
The writer process generally restarts from its beginning, while the reader process falls back to the last checkpoint, allowing the Teradata Access Module for Named Pipes to synchronize the two. You might need to take the following additional steps to complete the restart operation, depending on which client load and unload utilities are used:
- Prepare the writer process source for a clean start.
The FastExport utility, for example, uses a log table to determine that a task was interrupted. To start an interrupted FastExport job from its beginning, you must first drop the FastExport log table.The log table is specified by the LOGTABLE command in the FastExport job script.
- Modify the reader process job script. For example, in the FastLoad utility, do the following:
- Remove any statement that drops the table being loaded.
- Remove any statement that creates the table being loaded.
- Launch both the writer and the reader processes as described in Using the Teradata Access Module for Named Pipes. The Teradata Access Module for Named Pipes uses the fallback data file from the interrupted job to locate the restart position in the data stream from the writer process.