DSXchange

Delete records from Table X in one job.
In second job identify the inserts and updates and do them in two seperate jobs OR send it in with Upsert option.

Is there any error message in the log?

You can probably script it. After every table load, have an after job subroutine fire off a shell that cats the file and greps for the appropriate word that identifies the table. Look for particular warnings and if it exists, send out an email from the script itself.

There is a routine somewhere in the archives that I wrote that adds hours, minutes, seconds, even milliseconds to timestamp. Search for it.

Try either of the following for the sales column

sales column looks fine. Your first column "date" seems to be the culprit. Is that truely a date field? How are you converting that and whats your target column type?

Specify a reject link and analyze what records are the culprits. I am suspecting that your decimal field has more than just numbers.

I dont see a commit in your stored proc. Does this stored proc. work properly from command line? Add a commit at the end or instead of delete, do truncate and see what happens.

Search would have revealed all. Infact, there is an open topic by you on the very same error here.

You can use the Join stage, lookup stage and a few others. You said [b]output a row[b], which column is going to decide that, the first column or the second column? Never the less you can create two datasets and do a lookup keyed on 1st column for one dataset and 2nd for its copy. Depending upon the...

chulett wrote:No, the technique is perfectly valid. You build the hashed file with X keys and lookup on < X in the UV stage.

Lets hope the OP really built the hashed file with X keys and specified < X in the where predicate.
Dont worry about guessing, Ill do it for you

Welcome Aboard Please use the search facility. There have been few discussions on this topic before. Can you explain slow? Where does your Oracle sit? There will be network introduced again, right. 25 Million records is not that much. You can probably do a table export to a flat file, ftp the file o...

Because your loading a hashed file which will remove duplicates and hence defeat the purpose of Multirow result set return. Stick to ODBC stage.

or something like

For debugging purposes, change all date formats to YYYY-MM-DD. See what happens.