|
|
Archives of the TeradataForum
Message Posted: Thu, 31 Jul 2003 @ 20:28:42 GMT
Subj: | | Re: ETL process question |
|
From: | | rcifuentes |
What we do is to assign a "load number" to each file we load. Each loaded record has this number as a field (an INMOD routine adds the
load number to the record when loading) and gets into the destination table as a column.
Additionally we keep track of the final status of each load in a "load_log" table, wich includes date, time, file name, load number, job
name, etc ...
When something goes wrong, we can identify the load number from this load_log table, and delete records having that load number from the
affected tables.
For tables being updated we don't have safeguards, but based on the same schema, using some extra space, you can keep the 'old' values
for one night in additional columns. If something goes wrong, you can take the "old value" columns and restore the values on the affected
rows.
| |