Azure Stream Analytics: Multiple Windows JOIN generates too many rows

0
votes

I'm doing multiple calculations on multiple Windows. Each Window is a hopping window that outputs every second (with different window duration of course).

Then I have only one output query that joins my 5 windows with a DATEDIFF(second, a, b) BETWEEN 0 AND 1 for each joined window.

The problem is that this query generates way more rows than the input but it should generate the same amount. ( 8 rows from the input, 4002 rows from the query)

I can't find anything in the doc that's talking about this behavior.

UPDATE 1:

  • I found that decreasing JOIN DATEDIFF range, I have less events.

  • When using a range of 0-50ms in all my queries, I have 17 events but when comparing them to the 8 input events, they contain duplicates of some events and do not include all events so data is not complete.

  • My query steps are like that:

Window_N -> Stats_Window_N (SELECT FROM Window_N JOIN Window_N) -> All_Stats (SELECT FROM Window_N JOIN All Other Windows)

All using the same DATEDIFF range.

1
azureazure-eventhubazure-stream-analyticscomplex-event-processing

1 Answers

0
votes

What's in the join criteria? Reminds me of a Cartesian product when the join doesn't have the expected ON clause.

Could you give a bit of the literal syntax? Talking about it in the abstract may be too difficult to grasp.

Thanks, Jason