Question about DSGetLinkInfo routine

dfkettle · Post by **dfkettle** » Mon Jan 19, 2015 9:14 am

Can someone tell me if the DataStage routine 'DSGetLinkInfo' is supposed to work when called from a parallel job? Or does it only work when called from a server job? The reason I asked is because it's described in the Server Job Developer's Guide, but not in the Parallel Job Developer's Guide.

Thanks.

chulett · Post by **chulett** » Mon Jan 19, 2015 9:47 am

It's a BASIC function so you either need a Server job or a custom Before/After routine to call it, that or a Server Shared Container. That being said, you should be able to use it on a Parallel job.

dfkettle · Post by **dfkettle** » Mon Jan 19, 2015 9:53 am

Thanks. Someone here had a parallel job that was calling it from an After Job routine, and it didn't always work as expected (although it ran). I'll ask him to change it to a server job and see if that fixes it.

ArndW · Post by **ArndW** » Mon Jan 19, 2015 10:20 am

The BeforeJob and AfterJob calls are executed as BASIC calls in the server engine. There is no reason to re-write the parallel job.

chulett · Post by **chulett** » Mon Jan 19, 2015 10:28 am

Right, mentioned those as viable avenues. What about it didn't work as expected?

dfkettle · Post by **dfkettle** » Mon Jan 19, 2015 11:03 am

Sometimes it doesn't return the correct row count.

chulett · Post by **chulett** » Mon Jan 19, 2015 11:17 am

Can you narrow down the 'sometimes'? What stage are you trying to get counts from?

dfkettle · Post by **dfkettle** » Mon Jan 19, 2015 12:41 pm

chulett wrote:Can you narrow down the 'sometimes'? What stage are you trying to get counts from? ...

That's what we're trying to figure out. It happens infrequently, and we haven't been able to find any pattern yet that might help us understand the problem. It's retrieving the number of rows extracted by a DB2 stage in the same job.

ArndW · Post by **ArndW** » Mon Jan 19, 2015 1:24 pm

How long after the job finishes is the DSGetLinkInfo() called? Sometimes things work a bit asynchronously. Also, what partitioning is used for that job and link - anything special or different?

dfkettle · Post by **dfkettle** » Mon Jan 19, 2015 1:49 pm

It's in the same job (it's wrapped up in an "after-job" subroutine, called with 'DSJ.ME'). It always runs with the same number of partitions (two). The only thing different about the failed run was that we had DS_PXDEBUG turned on (for other reasons). Maybe that was just a coincidence, though.

stuartjvnorton · Post by **stuartjvnorton** » Mon Jan 19, 2015 5:16 pm

Do you find it mainly plays up when your InfoSphere server is getting worked hard?

I've had the situation where with not all of the individual node rowcounts make it back in time for the DSGetLinkInfo() call, resulting in discrepancies.
In our case, DataStage was set up for 8 nodes, the batch ran 8 simultaneous jobs, and it only happened with large, complex jobs.

Putting a short delay before the DSGetLinkInfo() call reduced the rate of occurrence greatly, but never quite eliminated it.

dfkettle · Post by **dfkettle** » Tue Jan 20, 2015 7:18 am

Thanks, you may be right. Like I said, it happened when we had debugging (DS_PXDEBUG) turned on. Maybe DataStage was busy.