[jira] [Commented] (BEAM-6958) Split Go PostCommit Test results by Runner

["Kyle Weaver (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/BEAM-6958?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16819515#comment-16819515 ] 

Kyle Weaver commented on BEAM-6958:
-----------------------------------

If I'm understanding correctly, at a high level, both the Java and Python rows run whatever they have designated "integration" tests for the SDK column [on the Dataflow runner|https://github.com/apache/beam/blob/9cb917569375b718af93f3ed2b9839e5629b647b/build.gradle#L145-L149] (meaning the existing Go SDK column actually already matches the others, as far as I can tell). The runner columns show results of whatever tests have been tagged as "validates runner" tests on their respective runners. So 2 questions:
 # Is this loose pattern established by Java/Python SDKs something worth emulating?
 # Do we have "validates runner" tests (or something similar) in Go?

The above notwithstanding, I think the most helpful thing for the Go SDK at this point might be to run the integration tests on Flink, just to prove we can. Thoughts?

> Split Go PostCommit Test results by Runner
> ------------------------------------------
>
>                 Key: BEAM-6958
>                 URL: https://issues.apache.org/jira/browse/BEAM-6958
>             Project: Beam
>          Issue Type: Improvement
>          Components: sdk-go, testing
>            Reporter: Robert Burke
>            Priority: Minor
>
> At present the Go SDK only has a single column filled in on the master branch Post-Commit Tests Status testing rubric, which is unclear, and not-ideal.
> Right now the jenkin's [Go PostCommit tests|https://github.com/apache/beam/blob/ec3f79214e9ef204fa32b744051a291fe4b61e23/.test-infra/jenkins/job_PostCommit_Go.groovy#L24] trigger the [go integration test task|https://github.com/apache/beam/blob/58a70b273367c22fd7c8562c42bc10a07dbe7156/build.gradle#L178], which only runs the [tests on Dataflow via a shell script|https://github.com/apache/beam/blob/master/sdks/go/test/run_integration_tests.sh#L78]. It doesn't even run the unit tests as per the pre-commit.
> The end goal for this task is to:
> * Have the Go SDK column represent the Go SDK Unit Tests as a post commit.
>   * Or better, to avoid pre-commit-run duplication, run the integration tests against the ULR if other runners are doing so.
> * Have the integration tests run against the Dataflow, be represented in the column.
> This will set the basis for adding and the integration tests against other portable runners (Flink, Spark, Python ULR, future portable runners...)
> It looks like there are three bits of work to accomplish here:
> * Adjust the Gradle tasks/task names to accurately represent what they're running against.
> * Add the new Jenkins tasks for each of the runners. (The other languages call these ValidateRunner_<RUNNER> tasks), 
> * Add the cool "badges" to the new Jenkins tasks to the Post Commit rubric.
>  



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)