Alternative to Sequence

PhilHibbs · Post by **PhilHibbs** » Thu Apr 17, 2008 4:14 pm

I don't really like sequences. Mainly my problem with them is they aren't restartable. Often I end up making a copy of a sequence and deleting the first half and running it from the point of failure.

I'd rather use something like a make script. Anyone had a go at using make scripts to run DataStage jobs? I don't really know make very well.

chulett · Post by **chulett** » Thu Apr 17, 2008 4:20 pm

Well, unless you are running something like the 7.0.1 version, you are going to get push-back on your "aren't restartable" comment. They are, we leverage that all functionality over the place. If you are on an older version, you can code in restart points (messy but doable) or you can... upgrade.

Of course there are alternatives, they've existed well before Sequence jobs themselves. You could always build your own 'job control' code or 'Batch::' jobs and do things as fancy there as you can whip up. Others have done this for you, for example Ken Bland gives away free his 'Job Control Utilities' which handle all of this rather neatly and easily, including restarting from any point in a series of jobs.

John Smith · Post by **John Smith** » Thu Apr 17, 2008 7:43 pm

You can always use the BASIC language and write a DS batch job. Include a FOR NEXT loop in there with parameterised FROM and TO, that way when you need to rerun you can choose which point to start and which point to end. In this case you will limit your run to a sequential run though.

ray.wurlod · Post by **ray.wurlod** » Thu Apr 17, 2008 8:28 pm

Not true (about limiting to sequential). You can build in as much parallelism as you desire, and you can start parallel jobs from a batch or a job control routine.

After all, a job sequence is only a GUI that generates a job control routine.

John Smith · Post by **John Smith** » Thu Apr 17, 2008 8:49 pm

You are right in that you can build as much parallelism into the batch job, but the OP requirement is for restartability from point of failure without using sequencers. So in the typical control job the jobs will be stepped through in sequence using a FOR loop.

The scenario provided by the OP was that he had to copy a sequence job and delete the first half which suggest to me it's sequential. That's where I was coming from.

So if batch job is running jobs 1 to 10 ,and job 5 aborted, then the restart has to be from job 5 till 10 and so on.

Now contrast this with in this same control job, I run in parallel all 10 jobs, and job 5 aborted then where do I restart?

ray.wurlod · Post by **ray.wurlod** » Thu Apr 17, 2008 9:20 pm

Depends which five. You have to build that intelligence in to your code.

chulett · Post by **chulett** » Fri Apr 18, 2008 6:39 am

Which Ken has already done.