understanding sequencer.sh to allocate nodes to job

[zulfi123786]

I was going through the Deploying Grid solution with Infosphere Information Server Redbook where a section describes the usage of sequencer.sh script to be run from within the sequence to create dynamic configuration file and use the same for jobs downstream.

The individual jobs however have different approach where osh_conductor.sh triggers Dynamic_Grid.sh which creates dynamic config file and holds resources and when the job finishes releases the resources back to resource manager.

In the sequencer if sequencer.sh is used only to return a dynamic configuration file, I presume the dynamic config file is created based on idle nodes at the time of creation but when the downstream jobs run how are resources held and released back as and when downstream jobs trigger and finish ?

[ray.wurlod]

Through interaction with the grid management software.

[PaulVL]

Ray, he's asking about how the place holder on the grid resource manager knows to be persistent until the end of the sequencer and not just till the end of the script that requested the nodes from the grid. All of the downstream jobs in that sequencer.sh job are apt_grid_enabled=no. because their apt file was pre-created. So no place holder in the manager during their execution.

Zul, I don't know. Personally, I'm not a big fan of the sequencer.sh tactic, but I do see where it can be useful at times. I tend to prefer having each job load balanced.

[zulfi123786]

Yes Paul, I too feel sequencer.sh is not a good option if the downstream jobs are bulky and the documentation also mentions the same that sequencer.sh to be used when the sequence has multiple light weight jobs to reduce the startup time which for light jobs might be comparable to run time.

As of now it looks better to load balance the individual jobs rather than call sequencer.sh and use the same dynamic config file.