Basic Teradata TPump Recovery
Whenever a database restart is detected or a Teradata TPump job is restarted on the
host system, the following activity occurs:
The restart log table is scanned with reference to the Teradata TPump script. Each
statement within the script is either executed because a row does not exist or ignored
because a row exists in the restart log.
In the case of the END LOAD statement, there are a number of rows which are placed
in the restart log table which let Teradata TPump decide what to do. Teradata TPump
ignores any complete IMPORT within a LOAD and begins at the incomplete IMPORT.
Within an unfinished IMPORT, Teradata TPump begins processing at the last complete
checkpoint. If the Teradata TPump job was running in SIMPLE mode before the restart,
then recovery is complete and processing continues at the last complete checkpoint.
If Teradata TPump was running in ROBUST mode before it was restarted, then Teradata
TPump must next ascertain how much processing has been completed since the last checkpoint.
This is accomplished by reading back a set of “Partial Checkpoints” from the restart
log table in Teradata Database, sorting them, and then reprocessing all transactions
which were left incomplete when the job was interrupted.