[ https://issues.apache.org/jira/browse/BEAM-122?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15300969#comment-15300969 ]
ASF GitHub Bot commented on BEAM-122: ------------------------------------- GitHub user mshields822 opened a pull request: https://github.com/apache/incubator-beam/pull/391 [BEAM-122][BEAM-311] Add GC hold if have data. Don't set timers beyond GlobalWindow.maxTim… R: @kennknowles You can merge this pull request into a Git repository by running: $ git pull https://github.com/mshields822/incubator-beam beam-122 Alternatively you can review and apply these changes as the patch at: https://github.com/apache/incubator-beam/pull/391.patch To close this pull request, make a commit to your master/trunk branch with (at least) the following in the commit message: This closes #391 ---- commit aed482670058ffc64ed37243a6ef5c4a21b2bd7c Author: Mark Shields <markshie...@google.com> Date: 2016-05-24T20:52:41Z Add GC hold if have data. Don't set timers beyond GlobalWindow.maxTimestamp ---- > GlobalWindow and allowedLateness can cause inconsistent timer interpretation > ----------------------------------------------------------------------------- > > Key: BEAM-122 > URL: https://issues.apache.org/jira/browse/BEAM-122 > Project: Beam > Issue Type: Bug > Components: runner-core > Reporter: Mark Shields > Assignee: Mark Shields > > In ReduceFnRunner we have code such as > window.getMaxTimestamp().plus(windowingStrategy.getAllowedLateness()) > If window is global then maxTimestamp will be > BoundedWindow.TIMESTAMP_MAX_VALUE. > Meanwhile, timestamps beyond BoundedWindow.TIMESTAMP_MAX_VALUE will be > clipped in most runners. > This could cause the time of an expected timer (eg for garbage collection) to > not match the actual time of a fired timer. > We should either make non-zero allowedLateness on the Global window illegal > (probably obnoxious) or explicitly clip it to zero. -- This message was sent by Atlassian JIRA (v6.3.4#6332)