Sequence starts job but then immediately ends?

RodBarnes · Post by **RodBarnes** » Thu Mar 19, 2009 4:16 pm

[I'm not even sure what I'd search for on this. If there is a known thread on this, please direct me.]

I have a sequence that invokes a job and, like any other sequence, when it does the sequence log reports:

Code: Select all

StatisticsLogs -> (jobImportVM): Job run requested
StatisticsLogs..JobControl (DSRunJob): Waiting for job jobImportVM to start

Typically, the sequence would wait and then this entry would show in the log:

Code: Select all

SessionLogs..JobControl (DSWaitForJob): Waiting for job jobImportVM to finish

Instead, rather than waiting for the job and then continuing on in the sequence, it immediately ends and the log shows:

Code: Select all

Finished Job StatisticLogs.

It reports no other errors or warnings. The sequence log gives no indication the job starts, fails, or anything else.

Checking the log for the job shows no errors and it actually completes its task successfully and imports the data I expected. The DSJobController correctly shows the name of the sequence. But any linkage to the sequence is broken.

Runing the job separately also completes successfully with no warnings or errors. I really am not sure where to begin to debug this.

ORACLE_1 · Post by **ORACLE_1** » Thu Mar 19, 2009 4:50 pm

1. Try to reset the job and see if that shows an error.

2. Go to the Job Control Tab in the Job Properties and see if there is a wait in there. Not sure if there is an option somewhere in the job to skip it or not , but that will give you an idea.

Thanks !

ray.wurlod · Post by **ray.wurlod** » Thu Mar 19, 2009 5:43 pm

Is jobImportVM a server job, a parallel job or a sequence job?

chulett · Post by **chulett** » Thu Mar 19, 2009 10:01 pm

I had something very similar happen with a Sequence job with a Start/End loop structure. It would enter the loop for the first time and show that the included Job Activity had been started, but then immediately exit not only the loop but the Sequence job as well. Logs looked just like you posted.

It was easily reproducible and made it all the way up to Level 2 Support before they gave up trying to explain the behaviour.

RodBarnes · Post by **RodBarnes** » Tue Mar 31, 2009 11:32 am

What did you end up doing to resolve it?
I am now seeing this in other jobs, too. Ones that worked fine -- and still do in earlier versions. For example, I have one sequence that still works fine but if I simply recompile the job in my work area, then re-import it to a test area, it will now exhibit this behavior. Yet, going back to the code in Version Control and promoting it, it works fine.

chulett wrote:I had something very similar happen with a Sequence job with a Start/End loop structure. It would enter the loop for the first time and show that the included Job Activity had been started, but then immediately exit not only the loop but the Sequence job as well. Logs looked just like you posted.

It was easily reproducible and made it all the way up to Level 2 Support before they gave up trying to explain the behaviour.

chulett · Post by **chulett** » Tue Mar 31, 2009 12:14 pm

I never could, really.

It happened with two jobs of almost identical design and once it started we never found a way to correct it, and "we" included both IBM and I. I ended up doing what I used to do before Sequence jobs were available - I hand-coded the equivalent looping process and didn't look back.

RodBarnes · Post by **RodBarnes** » Tue Mar 31, 2009 12:40 pm

This is really sad. Jobs which have worked fine for a long time now, when recompiled, will exhibit this. Yet a new job I just created to test a loop and promoted it forward, works in both locations. Clearly something has changed on the system or in DataStage but I haven't a clue where to begin to look. Next step is to try recreating one of those jobs from scracth and see if it still works.

chulett wrote:I never could, really.

It happened with two jobs of almost identical design and once it started we never found a way to correct it, and "we" included both IBM and I. I ended up doing what I used to do before Sequence jobs were available - I hand-coded the equivalent looping process and didn't look back.

RodBarnes · Post by **RodBarnes** » Tue Mar 31, 2009 2:19 pm

Well, in my case anyway, I've determined that it has to do with having a sequence stage be the immediate precursor to the startloop stage. If I eliminate the sequence stage and just change the two prior stages (which were joined at the sequence stage) to be inline to the startloop stage, it works fine.

Of course, the golden questions are: Why did it used to work and now doesn't? What changed?