Wave.sequence.error in phantom job

katz · Post by **katz** » Wed Apr 05, 2006 7:53 am

I am getting the following error during the last stage of my job (which happens to be an OCI) and my job aborts.

DSD.StageStatus: Wave sequence error, resource "testMEtemplate.Transform_and_Validate.OverWrite_RS_HF.SI_Pass_Thru_read" has CUR.WAVE.NUM=2, was looking for wave 1

Although though the error occurs in the last stage of the job, the stage referenced in the error occurs much earlier in the design. My guess is that the error is happening during an end of job clean-up process which happens to coincide with the execution of the final OCI stage. The final OCI stage performs an insert is not directly linked to the stage named in the error message (Overwrite_RS_HF)

The stage named in the error message is a transformer and the link that is referenced in the error (SI_Pass_Thru_read) is the primary input. The transformer is also has a lookup, but the primary input link is simply mapped to the output link - nothing fancy.

The job is also quite straight forward. Its not multi-instance enabled and uses a fairly routine set of stages (OCI, Hashed and Sequential Files, Transformers, and a Link Collector). Nothing in the job design violates any rules, that I am aware of.

I am unclear about what the error message means, so do not know how to address the problem in my job design. Does anyone know what this message means and what events can cause it?

kcbland · Post by **kcbland** » Wed Apr 05, 2006 8:21 am

You probably had an orphaned process from a previous failure. If you stop a job or it aborts, sometimes a part of a job that's talking to a database stays active. The next time the job is run, weird things happen if a thread is out there. Use "ps -ef |grep phantom" and look for parts of your job and kill them.

ray.wurlod · Post by **ray.wurlod** » Wed Apr 05, 2006 3:58 pm

Resetting or recompiling the job should correct any wave synchronization problems. A wave number is simply an internal DataStage mechanism for keeping track of unique runs - for example, resetting a job uses a new wave number because it can't be certain that the previous wave is in a runnable state (indeed, it probably isn't if you're resetting the job).

kcbland · Post by **kcbland** » Thu Apr 06, 2006 7:47 am

No, because the wave number gets reset all the time. Simply recompiling resets it to 1. Just ignore that it's there.