[jira] [Commented] (FLINK-7473) Possible Leak in GlobalWindows

["Steve Jerman (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/FLINK-7473?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16131745#comment-16131745 ] 

Steve Jerman commented on FLINK-7473:
-------------------------------------

OK. I think I have figured my issue.

When setting my timeout I used the ctx.getCurrentWatermark() as the base ... seems like this is set to max int... Changing to use just the event time stamp seems to fix the leak.



> Possible Leak in GlobalWindows
> ------------------------------
>
>                 Key: FLINK-7473
>                 URL: https://issues.apache.org/jira/browse/FLINK-7473
>             Project: Flink
>          Issue Type: Bug
>          Components: DataStream API
>    Affects Versions: 1.3.2
>         Environment: See attached project
>            Reporter: Steve Jerman
>         Attachments: timerIssue.zip
>
>
> Hi,
> I have been wrestling with a issue with GlobalWindows. It seems like it leaks instances of InternalTimer.
> I can't tell if it's a bug or my code so I created a 'minimal' project that has the issue...
> If you run the Unit Test  in the attached and then monitor heap you will see that the number of InternalTimers continually increases. I added code to explicitly delete them.. doesn't seem to help.
> If I comment out registerEventTimeTimer ... no leak :)
> My suspicion is that PURGE/FIRE_AND_PURGE is leaving the timer in limbo.
> Steve



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)