[
https://issues.apache.org/jira/browse/BEAM-12353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17350865#comment-17350865
]
Jan Lukavský commented on BEAM-12353:
-------------------------------------
Sure. The test validates that we are able to correctly apply change in WindowFn
after a GBK. It resulted from discussion about a change in SparkRunner, which
replaced `FullWindowedValueCoder` with `ValueOnlyWindowedValueCoder` (thus
dropping timestamps), which resulted in fail in this test, because all data was
placed into the same window. Long story short - this test validates that we
propagate timestamp correctly downstream from GBK.
> org.apache.beam.sdk.transforms.GroupByKeyTest$WindowTests.testRewindowWithTimestampCombiner
> is failng on Dataflow runner v2
> ---------------------------------------------------------------------------------------------------------------------------
>
> Key: BEAM-12353
> URL: https://issues.apache.org/jira/browse/BEAM-12353
> Project: Beam
> Issue Type: Bug
> Components: runner-dataflow
> Reporter: Boyuan Zhang
> Priority: P2
> Time Spent: 50m
> Remaining Estimate: 0h
>
> WindowTests.testRewindowWithTimestampCombiner is introduced recently by
> https://github.com/apache/beam/pull/14667. This test passes on Dataflow
> runner v2 both batch and streaming, but fails on Dataflow runner v2.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
