[ 
https://issues.apache.org/jira/browse/GEARPUMP-317?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16053233#comment-16053233
 ] 

ASF GitHub Bot commented on GEARPUMP-317:
-----------------------------------------

Github user manuzhang commented on a diff in the pull request:

    https://github.com/apache/incubator-gearpump/pull/187#discussion_r122594220
  
    --- Diff: 
streaming/src/main/scala/org/apache/gearpump/streaming/dsl/task/GroupByTask.scala
 ---
    @@ -61,13 +62,14 @@ class GroupByTask[IN, GROUP, OUT](
       }
     
       override def onWatermarkProgress(watermark: Instant): Unit = {
    -    groups.values.forEach(new Consumer[WindowRunner[IN, OUT]] {
    -      override def accept(runner: WindowRunner[IN, OUT]): Unit = {
    -        runner.trigger(watermark).foreach {
    -          result =>
    -            taskContext.output(Message(result.value, result.timestamp))
    +    if (groups.isEmpty && watermark == Watermark.MAX) {
    --- End diff --
    
    Other the `processingWatermark` will never advance when there are no 
incoming messages and `groups.isEmpty`


> Task minClock could be non increasing
> -------------------------------------
>
>                 Key: GEARPUMP-317
>                 URL: https://issues.apache.org/jira/browse/GEARPUMP-317
>             Project: Apache Gearpump
>          Issue Type: Bug
>          Components: streaming
>    Affects Versions: 0.8.3
>            Reporter: Manu Zhang
>            Assignee: Manu Zhang
>
> Task minClock is the minimum of upstream minClock and output minClock. 
> Considering the following scenario,
> 1. Output minClock is set to Long.MaxValue since all messages have been acked 
> by downstream tasks
> 2. But there are messages buffered in a window function
> 3. Later the window is triggered and those messages are emitted, which will 
> set the output minClock to a smaller value.
> 4. Plus, the output messages are possibly not in time order, which again will 
> make output minClock non increasing.



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

Reply via email to