DS_FORCE_ABORT_AT_WARN_LIMIT

[Raftsman]

We would like to force an abort if any warning errors are thrown from Datastage. The IBM documentation says that this utility will work. Seems not.

Has anyone ever used this and is there a specific way to implement. I tried it at the Job level, Sequence level and then the Project level with no success.

We could use the "dsjob -run -warn 1" but it will pass if only one warning is thrown.

Could someone please shed some light on this,

Thanks

[chulett]

We could use the "dsjob -run -warn 1" but it will pass if only one warning is thrown.

Actually, that means abort on the first warning.

[Raftsman]

Thanks, I also opened up a PMR asking about this trigger option. Not sure why it's published and doesn't work.

[ray.wurlod]

Isn't "as soon as the first warning is generated" the same, but more efficient, than "if any warnings are thrown"?

If you want to process all rows then abort if any warnings are thrown, you can implement an after-job subroutine to check the interim status (for example if that is not "Finished OK" then you can cause the job to abort).

[Mike]

I almost always run dsjob with a warning limit of 1 for production. As Craig says, the job will abort when the first warning is generated. It works most of the time. On occasion, I have observed that a single warning might get written to the log, but the job finishes too quickly after the warning and before DataStage has a chance to abort it.

That's generally not been an issue for me since the job will end with a "Finished (see log)" status and I have it set to generate a warning in the controlling job sequence which will then abort itself. I've never seen a case where a single warning made it past both a parallel job and a job sequence without aborting.

Mike

[chulett]

Thanks, I also opened up a PMR asking about this trigger option. Not sure why it's published and doesn't work.

I only found this documented in the 9.1 release... is that what you are running? And did you see in the documentation for it that they use "if" several times, meaning it may not always work depending on the timing.

If a parallel job produces enough warning messages to trigger the warning limit, an attempt is made to stop the job. If the job is successfully halted, the job termination status is set to ABORTED and the job must be reset before it can be run again. If the parallel job runs to completion before it can be stopped, the job status is set to FINISHED with WARNINGS because all of the input rows had been processed and there is no need for a reset.