Unlinked stages in one job

aaryabhatta · Post by **aaryabhatta** » Fri May 22, 2009 4:02 am

Can somebody help me understanding the process flow of a job when there are different inputs and write to the separate stages unlinked?
Like
dataset --> Oracle stage
dataset --> Oracle Stage
dataset --> Oracle stage all in the same job

Will all the inputs start processing at the same time sharing the nodes? How is it different from having three differnt jobs? Is there any advantage of jobs developed in this manner?

Thanks in Advance!

Regards

nagarjuna · Post by **nagarjuna** » Fri May 22, 2009 5:43 am

All these will start at the same time.Depending on the requirement If you want more control then go for a sequence .

chulett · Post by **chulett** » Fri May 22, 2009 6:57 am

As noted, they will all start at approximately the same time and run as if they were three separate jobs. However, a failure in one job stream could doom all of the others, depending on the timing. It could also complicate restarts.

No advantage, really, that I can think of. I'll guess that the thought was 'one job rather than three' (since three jobs must be evil) but to me it's just... lazy.