Records lost when reading a hashed file after upgrading DS

[NSIT-DVP-01]

Dear all,

We are testing the jobs after their migration from DataStage 7.5 to DataStage 8.7, and we are facing the following mismatch:

- there is one job that writes the records to the Hashed File - the same number of records is written by 7.5 and 8.7 versions of datastage.

- now the next job reads from that Hashed File, but DataStage 8.7 retrieves two records less than DataStage 7.5, from the exact same Hashed File, and with the same job design.

Does anyone have any idea how it is possible?
Is there a way to compare two hashed files to determine which 2 records get lost?

Additional info, if it might be important: we move from windows to unix server as well, and nls is enabled in the 8.7 version.

Thank you!

[manoj_23sakthi]

Hi,

If we need to check the 7.5 and 8.7 hased file. let we write in sequential file for both 7.5 and 8.7 hash file .we can compare both records.

Regards,
Manoj

[NSIT-DVP-01]

Thank you, I am trying this. I was thinking - may be it is related to the execution date used somewhere in the WHERE clause during the updates.

[ArndW]

Have you identified the 2 keys that work in the old version but don't in the new?

[chulett]

I was thinking it might be related to having NLS enabled in only the new version, some difference in the key values. Curious to know what it was...

[kduke]

You could have control characters in your key columns which might do it especially with NLS enabled on one and not the other.

You can select using a Pick style select and find the missing ones.

At TCL prompt

SELECT HashedFileName
SAVE-LIST HF1

The list is in a directory below the project called &SAVEDLISTS&.

Copy the HF1 to the &SAVEDLISTS& in the project on the new version.

GET-LIST HF1
NSELECT HashedFileName
SAVE-LIST Diff1