Different results in every run

srinivas.nettalam · Post by **srinivas.nettalam** » Mon May 27, 2013 2:42 pm

Hi All,
I developed a job with Oracle Connector as source ,then a lookup stage with 4 reference links out of which one link is a dummy lookup to populate a batch id and the other three come from a single oracle connector with a copy stage after that.The three reference links do not have any duplicates on the key column and the same reference key is looked up by three different stream keys.After the lookup I have plenty translations and transformations to do in transformer based on the lookedup values.When I run the job filtering a set of records (for e.g where itemid=xxx ) the output is as desired and when I remove the filter and run for the whole data the column values are scrambled.Only records with valid reference values are populated perfectly and others are populated with values which are not related to the field in anyway..When I run the job again ,the values populated in the previous run gets changed for the same columns though source data remains unchanged.

I tried separating the reference form copy stage to three individual reference stages and partitioning the data on the item field,the output was still not as expected.

Please suggest a way to troubleshoot this situation.

ray.wurlod · Post by **ray.wurlod** » Mon May 27, 2013 4:39 pm

How are your data partitioned on the inputs to the Lookup stage (assuming it's first after the Connector stage)?

srinivas.nettalam · Post by **srinivas.nettalam** » Tue May 28, 2013 1:21 am

Lookup is rght after the connector(source) .The connector is running in Sequential mode and I made Auto partition in lookup.When I tried parititioning the data on a field Item.I made hash partition on input and entire on the reference.The results were not as expected even in that case.

spoilt · Post by **spoilt** » Tue May 28, 2013 4:57 am

Use round robin after Source (which is sequencial) and Entire for every reference link.

srinivas.nettalam · Post by **srinivas.nettalam** » Tue May 28, 2013 8:29 am

I assume by default datastage assigns Round Robin to a stage following a stage running sequential mode.

srinivas.nettalam · Post by **srinivas.nettalam** » Thu Sep 05, 2013 11:17 pm

I rebuilt the job by separating the references to different look up stages and rewrote the stagevariables and constraints.It worked.