Thanks Timo for the suggestion! Also apologies for missing your response
last week. I will try to come up with a reproducible test case.
On Wed, Mar 18, 2020 at 9:27 AM Timo Walther wrote:
> Hi Vinod,
>
> thanks for answering my questions. The == Optimized Logical Plan ==
> looks as expected.
Hi Vinod,
thanks for answering my questions. The == Optimized Logical Plan ==
looks as expected. However, the == Physical Execution Plan == seems to
be quite complex. Are you sure that watermarks don't get lost in some of
those custom operators before entering the SQL part of the pipeline?
I wanted to add that when I used the following the watermark was delayed by
3 hours instead of 2 hours that I would have expected:
AND o.rowtime BETWEEN c.rowtime - INTERVAL '2' hour AND c.rowtime
(time window constraint between o and c: 1st and 3rd table)
Thanks,
Vinod
On Fri, Mar 13, 2020 at
Thanks Timo for responding back! Answers below:
> 1) Which planner are you using?
We are using Flink 1.8 and using the default planner
(org.apache.flink.table.calcite.FlinkPlannerImpl)
from: org.apache.flink:flink-table-planner_2.11:1.8
> 2) How do you create your watermarks?
We are using
Hi Vinod,
I cannot spot any problems in your SQL query.
Some questions for clarification:
1) Which planner are you using?
2) How do you create your watermarks?
3) Did you unit test with only parallelism of 1 or higher?
4) Can you share the output of TableEnvironment.explain() with us?
Hi!
We are testing the following 3 way time windowed join to keep the retained
state size small. Using joins for the first time here. It works in unit
tests but we are not able to get expected results in production. We are
still troubleshooting this issue. Can you please help us review this in