DSXchange

Splitting your stream into two - Insert only and Update only - is the usual recommendation to achieve shortest execution time. Upsert works by trying one (whichever you have specified as happening first) then, if that fails because of the primary key constraint, trying the other. You can avoid the t...

If you search DSXchange you will find that Maintenance mode does render indexes unusable - you need to disable SKIP_INDEX_MAINTENANCE or a Close command to remedy that.

1. Not really. Multiple readers is about all you have unless you chop up the file first. 2. No. Sparse lookups are only available for DB2 and Oracle Enterprise stages. There is a virtual Data Set associated with very other link between non-combined operators. You *may* get some gain by preventing th...

But the original works OK - you just get row counts totalled across all partitions only.

No. I am talking about the Field Width property. Right click on a row in the Columns grid and choose Edit Row from the menu. You'll find it there. (The Field Width *may* show up as the Display Width in the Columns grid, but definitely not as the Length. Length is the maximum size of a VarChar.)

"Maintenance". Because it works.

I think the question is whether ETLStats can gather data from parallel jobs, and the answer to that question is "yes".

Is the CFF stage (operator) combined with the Copy stage?

If so, there is only one operator from which row counts can be obtained, and these are usuallly displayed on the output of that operator - the most downstream link.

Try disabling operator combination to see what I mean.

If you go to IOD 2008 in Las Vegas (as an attendee) you can take a certification preparation class (from memory it's about $400) and then attempt the exam itself for free. Check out the IOD web site for more information. Of course there are no guarantees - you will definitely not pass if you go into...

The manuals are called odbcref.pdf and odbcuser.pdf (reference guide and user guide respectively) - and, for version 7.5x2, they are the first two links from page 2 of the bookshelf.

dspxguy wrote:i had to use my laptop and it had vista...

I suspect you/they have violated your client's end user licensing agreement by requiring that.

node is a unique name for each (logical) processing name; it is used primarily in reporting error messages - internally DataStage uses node #0, node #1 and so on. fastname is the network node name of the machine on which that processing node will execute. If you are only using one machine the value ...

Unless you tell us precisely how you are processing in the Transformer we can not answer your question. It IS possible for more rows to come out of a Transformer than go in, but you have to program this - it can not happen of its own accord.

Not in source, only in target. Derive as MIN(XX_MA_DT) and MAX(XX_MA_DT) respectively.

Yes, that seems likely. You need to make sure that column #138 contains all the characters needed by the target database.