DSXchange

The most important factor influencing that decision is: what is the ratio between inserts and updates? It will be faster to split the stages into 2 distinct operations as pure inserts are generally faster and often using 2 connections to the database as opposed to just 1 can improve throughput.

Andy, I'm at a SQL-Server site at present and have to deal with sporadic problems and aborts a lot - with no resolution in sight. In my case we are using PX jobs, though. Is the system heavily loaded when this happens, and/or do a lot of SQL-Server connections get established within a short time fra...

Odd indeed. I can't check now, but there are a number of additional logging parameters you could set including APT_DUMP_SCORE, and then variables something like 'report player calls' and 'report process ids', etc. Perhaps that might help you narrow down exactly which call from which process is the l...

This is a good example of where a dataset should be used.

Could you put some CALL DSLogInfo() calls into your before-job routine, particularly one right before the RETURN statement? That will ensure that the cause of the hang is in the DataStage job and not in the before-job code.

That is standard japanese text, discussing host downsizing.

Thanks Craig, I missed that part! I was using a target stage and couldn't see any problems with a reject stage, but we don't have DB2 here so I couldn't actually run a job.

I don't know what could trigger a reject on a source, either. I don't know if that is possible to to do in DataStage.

The "@ID" is the key column of a hashed file. The "F1" is shorthand for "Field 1". Thus SELECT all rows in <hash_file_name> where the key is not an empty string and sort by ascending column 1. This column can be seen by "LIST DICT <hash_file_name>" and finding...

There should be one output link from the DB2 stage and it should be of type "reject", i.e. a dotted line.
Is this a compile-time error or a runtime error?

That's quite odd, I did a test job earlier to see about the aggregator and had a functioning max() working using a sort and remove duplicates within 5 minutes. Perhaps you might want to revist your approach, and if you wish assistance here you might wish to post what you did and what the problem was.

I just tried it and it would seem that the MAX() function of the aggregator stage is limited to numeric data in Parallel jobs. Along similar lines to what Ray mentioned, use one stream for a sort and a remove duplicates (first or last, depending upon whether you want a min or a max value) to get the...

Sometimes multinstance jobs and director log display don't get along and an instance of such a job might show empty. Is the job in question a multi-instance one?

Call me a fetishist, but I like generic jobs - easier to invest time to make one job handle 100 inputs/outputs than write 100 job copies to do the same (and later on have to make a small change to 100 jobs...). This case is a bit more difficult than a generic generic job, though. You would need to p...

The Job Monitor sometimes shows all "green" even when the job has aborted and is not to be trusted to always show this value correctly.

Is the job log for the aborted job always empty or just empty after an abort event?

DSX.UTILITYHASHLOOKUP is a function which returns a non-zero value at runtime, when used directly in your job sequence it will trigger that warning message.

How about assigning the return value to a user variable in a user variables activity to avoid this warning?