DSXchange

You'd be surprised how much data you can shift with server jobs, particularly if you use the multi-instance capability. While this can't partition processing across an MPP cluster, it can take full advantage of multiple CPUs in an SMP environment. The only downside, compared to PX, is that you (the ...

My guess is that there's something awry with the data in the column `asset`.`residpct` in that row. Run the job in debugger; set a breakpoint on the input link and edit its expression to @INROWNUM >= 34998. Execution of the job will pause at row number 34999, and you can then step through at human s...

There's no context switching when calling a routine; the only overhead is determining its location (via the Catalog, or VOC file) and loading it into memory (this process is sometimes called "link snapping"); after that, the in-memory location is cached. As part of loading the routine, stack entries...

I occasionally see this on DS 7.0.1.1r2 on AIX, in quite simple job designs. Haven't yet tracked down a likely cause, but pretty sure it's not job complexity. Think it's a bug.

My guess would be that you might be running out of hashed file descriptors in the T30FILE table occasionally, or maybe just taking a timeout accessing the resource file. That said, I don't think it could be the latter, because this is a key-based access to a hashed file that is never updated and the...

The only way that DataStage (or you, or anything else for that matter) could do it is to make reference to the target Oracle table, to determine whether the key exists already. If it does, do an update. If it doesn't, do an insert. Given that there is a key, this is all that you need to load into th...

That's easy. If you have one processing node, then you have only one processing node, and no partitioning of the data will take place. The round robin algorithm will place the first row onto node 1 (of 1), then the second row onto node 1 (of 1), and so on.

Since you've now specified BASIC I will assume you're using a server routine, not anything from the parallel environment, and answer accordingly. Declare a dimensioned array with the Dimension statement. For example: Dimension MonthlyRainfall(12,3) These elements are in an unassigned state; ...

This one may be faster. FUNCTION PreviousWeekdayToDate(TheDate, DayName) * TheDate is expected to be internal date format. * By using EQUATE, concatenation is performed at compile time. Equate DayList To "SUNDAY" : @FM : "MONDAY" : @FM : "TUESDAY" : @FM : "...

"Unable to remove PRINTER segment" suggests that the permissions on the uv and/or dsapi_slave executable are wrong. These must be owned by root and have the setuid bit set. These commands execute an ipcrm (as root) before switching to the actual user to invoke the shell (for example uvsh ). If you d...

You need to re-think the design.

Have one link doing insert only, and another link doing update only.
Make the decision by reference lookup against the target table or - ideally - a copy of the relevant rows and columns from it in a hashed file.

You will be amazed at the difference this makes!

Nice thought, but JOINBUF is only about providing scratch space for optimizing joins in "UniVerse SQL". The major tip for optimizing DataStage jobs is the word "don't". Don't process any rows you don't have to. Don't process any columns you don't have to. Don't make expressions or routines unnecessa...

Can we just confirm whether you're really using parallel (as your post states) or server job? The answer is radically different, depending on which kind of job you're using. Also, did you intend to use a "conventional" dimensioned array, or a dynamic array (with an arbitrary number of arbirarily-siz...

About the only way is to use a UniData stage to retrieve a single column, and that is defined as an I-descriptor whose expression is @RECORD (or, if you need the key, @ID:@FM:@RECORD).

No. As soon as you call DSLogFatal that job aborts immediately. Therefore any after-stage or after-job subroutine can not be executed. That's why we recommend not to use this function. The only way you can detect an aborted job (and, possibly, reset it) is in its controlling job. Why not use an OCI ...