You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@beam.apache.org by "ASF GitHub Bot (JIRA)" <ji...@apache.org> on 2018/12/11 17:50:00 UTC

[jira] [Work logged] (BEAM-3863) AfterProcessingTime trigger doesn't fire reliably

     [ https://issues.apache.org/jira/browse/BEAM-3863?focusedWorklogId=174180&page=com.atlassian.jira.plugin.system.issuetabpanels:worklog-tabpanel#worklog-174180 ]

ASF GitHub Bot logged work on BEAM-3863:
----------------------------------------

                Author: ASF GitHub Bot
            Created on: 11/Dec/18 17:49
            Start Date: 11/Dec/18 17:49
    Worklog Time Spent: 10m 
      Work Description: mxm commented on issue #4875: BEAM-3863: AfterProcessingTime trigger firing at delayedUntil time
URL: https://github.com/apache/beam/pull/4875#issuecomment-446295903
 
 
   It seems like this could be easily fixed by correctly converting the Beam timestamp into a Flink timestamp.

----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
 
For queries about this service, please contact Infrastructure at:
users@infra.apache.org


Issue Time Tracking
-------------------

    Worklog Id:     (was: 174180)
    Time Spent: 2h  (was: 1h 50m)

> AfterProcessingTime trigger doesn't fire reliably
> -------------------------------------------------
>
>                 Key: BEAM-3863
>                 URL: https://issues.apache.org/jira/browse/BEAM-3863
>             Project: Beam
>          Issue Type: Bug
>          Components: runner-flink
>    Affects Versions: 2.1.0, 2.2.0, 2.3.0
>            Reporter: Pawel Bartoszek
>            Assignee: Aljoscha Krettek
>            Priority: Major
>          Time Spent: 2h
>  Remaining Estimate: 0h
>
> *Issue*
> Beam AfterProcessingTime trigger doesn't fire always reliably after a configured delay.
> The following job triggers should fire after watermark passes the end of the window and then every 5 seconds for late data and the finally at the end of allowed lateness.
> *Expected behaviour*
> Late firing after processing time trigger should fire after 5 seconds since first late records arrive in the pane.
> *Actual behaviour*
> From my testings late triggers works for some keys but not for the other - it's pretty random which keys are affected. The DummySource generates 15 distinct keys AA,BB,..., PP. For each key it sends 5 on time records and one late record. In case late trigger firing is missed it won't fire until the allowed lateness period. 
> *Job code*
> {code:java}
> String[] runnerArgs = {"--runner=FlinkRunner", "--parallelism=8"};
> FlinkPipelineOptions options = PipelineOptionsFactory.fromArgs(runnerArgs).as(FlinkPipelineOptions.class);
> Pipeline pipeline = Pipeline.create(options);
> PCollection<String> apply = pipeline.apply(Read.from(new DummySource()))
>         .apply(Window.<String>into(FixedWindows.of(Duration.standardSeconds(10)))
>                 .triggering(AfterWatermark.pastEndOfWindow()
>                         .withLateFirings(
>                                 AfterProcessingTime
>                                         .pastFirstElementInPane().plusDelayOf(Duration.standardSeconds(5))))
>                 .accumulatingFiredPanes()
>                 .withAllowedLateness(Duration.standardMinutes(2), Window.ClosingBehavior.FIRE_IF_NON_EMPTY)
>         );
> apply.apply(Count.perElement())
>         .apply(ParDo.of(new DoFn<KV<String, Long>, Long>() {
>             @ProcessElement
>             public void process(ProcessContext context, BoundedWindow window) {
>                 LOG.info("Count: {}. For window {}, Pane {}", context.element(), window, context.pane());
>             }
>         }));
> pipeline.run().waitUntilFinish();{code}
>  
> *How can you replicate the issue?*
>  I've created a github repo [https://github.com/pbartoszek/BEAM-3863_late_trigger] with the code shown above. Please check out the README file for details how to replicate the issue.
> *What's is causing the issue?*
> I explained the cause in PR.
>  



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