flink-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Tzu-Li (Gordon) Tai (JIRA)" <j...@apache.org>
Subject [jira] [Reopened] (FLINK-7728) StatusWatermarkValve has different min watermark advancement behavior depending on the ordering inputs become idle
Date Mon, 12 Mar 2018 15:25:01 GMT

     [ https://issues.apache.org/jira/browse/FLINK-7728?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

Tzu-Li (Gordon) Tai reopened FLINK-7728:
----------------------------------------

> StatusWatermarkValve has different min watermark advancement behavior depending on the
ordering inputs become idle
> ------------------------------------------------------------------------------------------------------------------
>
>                 Key: FLINK-7728
>                 URL: https://issues.apache.org/jira/browse/FLINK-7728
>             Project: Flink
>          Issue Type: Bug
>          Components: DataStream API
>    Affects Versions: 1.2.1, 1.4.0, 1.3.2
>            Reporter: Tzu-Li (Gordon) Tai
>            Assignee: Tzu-Li (Gordon) Tai
>            Priority: Blocker
>             Fix For: 1.4.0, 1.3.4
>
>
> Currently, once all inputs of a {{StatusWatermarkValve}} becomes idle, we only emit the
{{StreamStatus.IDLE}} marker, and check nothing else. This makes the watermark advancement
behavior inconsistent in the case that all inputs become idle, depending on the order that
they become idle.
> Consider the following setup:
> {code}
> Channel #1: Watermark 10, ACTIVE
> Channel #2: Watermark 5, ACTIVE
> Channel #3: Watermark 5, ACTIVE
> {code}
> If the channels become IDLE with the order #2 -> #3 -> #1, watermark 10 will be
emitted as the new min watermark once both #2 and #3 becomes idle.
> On the other hand, if the order is #1 -> #2 -> #3, watermark 10 will not be emitted.



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

Mime
View raw message