[
https://issues.apache.org/jira/browse/APEXCORE-375?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15182522#comment-15182522
]
Vlad Rozov commented on APEXCORE-375:
-------------------------------------
The issue is caused by catchUp being scheduled to run while actual catchUp may
be invoked when DataList notifies its listeners. The scheduled catchUp will be
executed unconditionally and may skip tuples till it gets begin window tuple.
This leads to payload and end window tuples not being delivered to downstream
containers.
> Container killed because of Out of Sequence tuple error.
> --------------------------------------------------------
>
> Key: APEXCORE-375
> URL: https://issues.apache.org/jira/browse/APEXCORE-375
> Project: Apache Apex Core
> Issue Type: Bug
> Affects Versions: 3.2.0, 3.3.0, 3.2.1, 3.4.0, 3.3.1
> Reporter: Tushar Gosavi
> Assignee: Vlad Rozov
>
> Under high load with MxN partitioning, if upstream operator is killed then
> operators unifiers (located at downstreams containers) gets killed randomly
> because of out of sequence tuples.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
