[
https://issues.apache.org/jira/browse/BEAM-1620?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16056835#comment-16056835
]
Kenneth Knowles edited comment on BEAM-1620 at 6/21/17 2:01 AM:
----------------------------------------------------------------
For streaming vs batch as controlled by PipelineOptions for e.g.
DataflowRunner, these are actually totally separate configs that can have
fairly granular exclusions that are decoupled. So maybe there are new
categories we might want but for this example it is easy to have the streaming
ValidatesRunner execution include UsesSplittableDoFn while the batch execution
excludes it.
was (Author: kenn):
I think we might already have a category for tests that use unbounded data.
That is the distinction we need.
For streaming vs batch as controlled by PipelineOptions, these are actually
totally separate configs that can have fairly granular exclusions that are
decoupled. So maybe there are new categories we might want but for this example
it is easy to have the streaming ValidatesRunner execution include
UsesSplittableDoFn while the batch execution excludes it.
> Add streaming Dataflow ValidatesRunner coverage
> -----------------------------------------------
>
> Key: BEAM-1620
> URL: https://issues.apache.org/jira/browse/BEAM-1620
> Project: Beam
> Issue Type: Test
> Components: runner-dataflow, testing
> Reporter: Kenneth Knowles
> Assignee: Jason Kuster
>
> Currently, the runner validation test suite is not run on Dataflow in
> streaming mode. In fact, it should be able to run - all the functionality is
> in place. I think this is just a matter of maven + Jenkins + making sure not
> to leak a bunch of streaming jobs.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
