Aggregator stranges

[snassimr]

Hi !

I meet sometimes starnge behavior of aggreagtor.
I have a simple job :

DB2 -> TRANSFORMER -> AGGREGATOR -> SEQFILE

In DB2 I have 1000000 rows when the job run statistics of rows in link between DB2 and TRANSFORMER rich 3000000 rows (3 times exactly f the source number rows)

Any explanation

[loveojha2]

Hi !

I meet sometimes starnge behavior of aggreagtor.
I have a simple job :

DB2 -> TRANSFORMER -> AGGREGATOR -> SEQFILE

In DB2 I have 1000000 rows when the job run statistics of rows in link between DB2 and TRANSFORMER rich 3000000 rows (3 times exactly f the source number rows)

Any explanation

What is the source query that you are running in DB2?
May be there are duplicates within the source table,
most probably it is result of a cartesian within the source query.

[snassimr]

Its not the case .

After the job finished the counter retutned to 1000000 but I feel the DS do more work that reqiered for 1000000 rows

[ArndW]

If you run the job with just 1 row input do you get 3 rows output? This will make your debugging a lot easier. Also, what are you doing in the aggregator stage?

[ray.wurlod]

I have also seen something like this behaviour. It also occurs in the job monitor, so is probably a problem with how the link row count is being collected when there is a direct link from a Transformer to an Aggregator stage. As noted, the final row counts are correct. It might be an artifact of row buffering - I never pursued it since I was only interested in the final row counts. Does someone have lots of free time to construct a duplicate case? (It doesn't need to be as many as 1 million rows, but does need to be a substantial number.)