DSXchange

An option is to use a Server job, the Sequential File stage there is more accomdating of issues like this for some reason.

Make sure the key columns are right at the 'DataStage level' so you have a proper where clause, any 'keys' in the actual target table only help it to go faster.

Not much information to go on... perhaps an update is clearing the field after the fact? Anything special about the 'some rows' that have this issue?

Concatenation with a space between. That plus either IConv/OConv or substring to convert the date to YYYY-MM-DD format.

Another option - writing to a Folder stage in a Server job.

Yup.

One methodology - DATE datatypes can be subtracted to get a difference in days, then you can check to see if that difference is > 730.

Hard to say from so little information, but typically a 'blink' error from pushing too many messages into it and causing it to grow past the 32bit addressing boundary of ~2GB. Solution? Don't allow unlimited growth.

I've never used the stage but can't imagine that would be true, what makes you think that? Something in the docs?

With checkpointing turned on in the 'child' sequences, you really don't want them to be reset as that will remove any checkpoints and rerun them from the beginning. I assume you want them to pick up from the aborted step, yes?

How exactly does 'the master sequence break again'?

It's called "job security" if you make it seem really, really complex.

Three targets with the appropriate action in each and constraints based on the last column's value.

... plus 'R' means Replace and the '30' is the numeric type of the hashed file to create, i.e. a dynamic hashed file.

You post images here the same way you do on any other web board - host the image somewhere and then link to it here. There are plenty of file file hosting sites and most will automatically build the img 'image' tags you'll need. The suggestions will come once people understand what it is you are try...

What does the documentation say about it?