[jira] [Comment Edited] (BEAM-6011) Enable Phrase triggering in Nexmark jobs

["Lukasz Gajowy (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/BEAM-6011?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16678445#comment-16678445 ] 

Lukasz Gajowy edited comment on BEAM-6011 at 11/7/18 4:15 PM:
--------------------------------------------------------------

I made some attempts to prepare a quick fix for this. It was not so easy as I initially expected. I tried to use the $GIT_BRANCH environment variable to conditionally create the nexmark args needed for big query connection. It seems that this is not the right approach, because $GIT_BRANCH is evaluated while running the job whereas the rest of the code is evaluated while running the Seed job. Moreover, I don't like this solution because it's hiding the real intent of the job.

Link to the PR (currently closed): [https://github.com/apache/beam/pull/6974]

IMO, after some trying, the way it has to be done is that: 
 - every nexmark job should have a "twin" job that is for running on PRs (exclusively)
 - PR triggered jobs should save to the big query database the same way as the periodic jobs (because saving also can change and we will need to test that in PRs too).

 - PR triggered jobs should save to a new BQ dataset and have the same table names.  

 - Jenkins dsl code should be refactored so that it provides convenient methods for creating new jobs =

CC: [~kenn] , [~echauchot] WDYT? Do you have other comments/suggestions?

We should do exactly the same for IO Performance tests (currently they suffer the same problems)...

 

 


was (Author: łukaszg):
I made some attempts to prepare a quick fix for this. It was not so easy as I initially expected. I tried to use the $GIT_BRANCH environment variable to conditionally create the nexmark args needed for big query connection. It seems that this is not the right approach, because $GIT_BRANCH is evaluated while running the job whereas the rest of the code is evaluated while running the Seed job. Moreover, I don't like this solution because it's hiding the real intent of the job.

Link to the PR (currently closed): [https://github.com/apache/beam/pull/6974]

IMO, after some trying, the way it has to be done is that: 
 - every nexmark job should have a "twin" job that is for running on PRs (exclusively)
 - PR triggered jobs should save to the big query database the same way as the periodic jobs (because saving also can change and we will need to test that in PRs too).

 - PR triggered jobs should save to a new BQ dataset and have the same table names.  

 - Jenkins dsl code should be refactored so that it provides convenient methods for creating new jobs =

CC: [~kenn] , [~echauchot] 

We should do exactly the same for IO Performance tests (currently they suffer the same problems)...

 

 

> Enable Phrase triggering in Nexmark jobs
> ----------------------------------------
>
>                 Key: BEAM-6011
>                 URL: https://issues.apache.org/jira/browse/BEAM-6011
>             Project: Beam
>          Issue Type: Bug
>          Components: testing
>            Reporter: Lukasz Gajowy
>            Assignee: Lukasz Gajowy
>            Priority: Critical
>
> We need to enable Phrase Triggering (running Jenkins jobs from PR) for Nexmark jobs so that we could check if pull requests are not breaking anything before merging them. 
> Note: Currently Nexmark jobs run post commit on master and publish their results to BigQuery database. In order not to pollute the results collected for master we should save the results for Pr-triggered jobs in some other tables/datasets or even not save them at all (turn off publishing to BQ).



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