Mailing-List: contact commits-help@beam.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@beam.apache.org
Date: Wed, 28 Dec 2016 22:21:58 +0000 (UTC)
From: "Davor Bonaci (JIRA)" <jira@apache.org>
To: commits@beam.incubator.apache.org
Message-ID: <JIRA.13001182.1472529356000.609194.1482963718467@Atlassian.JIRA>
In-Reply-To: <JIRA.13001182.1472529356000@Atlassian.JIRA>
References: <JIRA.13001182.1472529356000@Atlassian.JIRA> <JIRA.13001182.1472529356070@arcas>
Subject: [jira] [Updated] (BEAM-604) Use Watermark Check Streaming Job
 Finish in TestDataflowRunner
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit
archived-at: Wed, 28 Dec 2016 22:22:02 -0000


     [ https://issues.apache.org/jira/browse/BEAM-604?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Davor Bonaci updated BEAM-604:
------------------------------
    Component/s: runner-dataflow

> Use Watermark Check Streaming Job Finish in TestDataflowRunner
> --------------------------------------------------------------
>
>                 Key: BEAM-604
>                 URL: https://issues.apache.org/jira/browse/BEAM-604
>             Project: Beam
>          Issue Type: Improvement
>          Components: runner-dataflow
>            Reporter: Mark Liu
>            Assignee: Mark Liu
>            Priority: Minor
>             Fix For: 0.3.0-incubating
>
>
> Currently, streaming job with bounded input can't be terminated automatically and TestDataflowRunner can't handle this case. Need to update TestDataflowRunner so that streaming integration test such as WindowedWordCountIT can run with it.
> Implementation:
> Query watermark of each step and wait until all watermarks set to MAX then cancel the job.
> Update:
> Suggesting by [~peihe0@gmail.com], implement checkMaxWatermark in DataflowPipelineJob#waitUntilFinish. Thus, all dataflow streaming jobs with bounded input will take advantage of this change and are canceled automatically when watermarks reach to max value. Also Dataflow runners can keep simple and free from handling batch and streaming two cases.
> Update:
> Pipeline author should have control on whether or not canceling streaming job and when. Test framework is a better place to auto-cancel streaming test job when curtain conditions meet, rather than in waitUntilFinish().


--
This message was sent by Atlassian JIRA
(v6.3.4#6332)