Archives of the TeradataForum

Message Posted: Sat, 03 Sep 2016 @ 14:46:06 GMT

	<Prev	Next>		<<First	<Prev	Next>	Last>>

Sreedhar,

When you run thousands of load jobs, it is not a bad idea to have restart logic built in to your job. It's easy enough to ignore an error message --- you can figure that out --- but you might want to consider one other thing.

The release mload command waits for an exclusive lock on the table. That can be an issue in a data warehouse environment with a lot of analysts executing long-running dirty read queries. And 99 out of 100 times (or more) you won't need to run the release mload statement.

So we first check to see if the work tables created by a previous load exist. (You should have strict naming conventions for the work tables so that you can detect that.) If they do, then the assumption is that the previous run failed and we need to not only release mload, but also drop the work and error tables. If, on the other hand, the work tables do not exist, we skip the release mload statement and avoid possible lock wait delays.

You have a good idea. Good luck!

Pat McKeown

	<Prev	Next>		<<First	<Prev	Next>	Last>>