DSXchange

We are facing an issue lately for a particular job which says 'Abnormal termination of stage tfmr detected' and I don't see any vital info from the previous run messages. All it says is that, which stage is using what NLS map.

E.g:
<job_name>.<stage_name>: Using project default NLS map ISO8859-1
<job_name>.<stage_name>: Using project default NLS map ISO8859-1
<job_name>.<stage_name>: Using project default NLS map ISO8859-1


I have searched and have checked that there are many posts on this matter here and I tried few things too like setting of DS_NO_PROCESS_METADATA=1. But there is not resolution for this. any help on this will be appreciated.

It looks like you will have to debug this the hard way. I would remove all columns that do a 1-1 mapping, then recompile and text. Chances are very high the error will still be there. Then start removing single columns, starting with the most complicated derivations first until the error goes away and then examine (or post to this thread) the offending derivation.

Do you think it is something related to a derivation? I doubt so. The reason is that this transformer's output is being fed into a Sort stage which eventually goes to another transformer as the primary link while we have an ODBC stage for connecting to Oracle that is used as reference for the 2nd transformer.

I replaced all the stages after the 1st transformer with a sequential file and executed the job. It ran successfully.

In that case you need to follow the same approach, but at a stage level instead of at a column level. Remove stages until the job runs, then the last stage removed is triggering the problem and you can drill down from there.

Do you think it is something related to a derivation? I doubt so. The reason is that this transformer's output is being fed into a Sort stage which eventually goes to another transformer as the primary link while we have an ODBC stage for connecting to Oracle that is used as reference for the 2nd transformer.

I replaced all the stages after the 1st transformer with a sequential file and executed the job. It ran successfully.

I got the same issue for one of my ETL jobs. After analysis I tweaked the array size used in the target OCI stage. Just try reducing the array size. It worked for my job. I hope it helps.

Please elucidate how this is the same issue - how a problem whose symptom is NLS map related can be fixed by tweaking array size.

I do not know of any way that this is possible, so I seek your enlightened guidance.

Ray,

This is what happened with me during one of the job executions. I resolved it by tweaking the array size. I don't know whether this will enlighten your thoughts or not .

I can confirm that the array-tweak approach does indeed fix the issue.

/Nick

It is possible if the target stage action is set to update then insert and the array size is greater than 1, which it should not be in this scenario to begin with, it will cause an abnormal termination error. Nobody on this topic has provided enough detail to say if that was truly the cause or not.

If it is the cause, then was the target stage actually named "tfmr?" Or was the error reportedly coming from a Transformer stage, and actually generated by the target database stage?

When Server jobs blow up, the "active" stage always reports the error... meaning that would typically be the transformer.