You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@flink.apache.org by "Yuan Mei (Jira)" <ji...@apache.org> on 2021/08/12 02:26:00 UTC
[jira] [Closed] (FLINK-23475) Handle BroadcastState with finished
operators
[ https://issues.apache.org/jira/browse/FLINK-23475?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Yuan Mei closed FLINK-23475.
----------------------------
Fix Version/s: 1.14.0
Resolution: Fixed
merged commit [{{1c563a0}}|https://github.com/apache/flink/commit/1c563a03e95a7baf91055ce1fec3d2a6f35c393d] into apache:master [10 minutes ago|https://github.com/apache/flink/pull/16714#event-5146140293]
> Handle BroadcastState with finished operators
> ---------------------------------------------
>
> Key: FLINK-23475
> URL: https://issues.apache.org/jira/browse/FLINK-23475
> Project: Flink
> Issue Type: Sub-task
> Components: Runtime / Checkpointing
> Reporter: Dawid Wysakowicz
> Assignee: Yun Gao
> Priority: Major
> Labels: pull-request-available
> Fix For: 1.14.0
>
>
> BroadcastState assumes all subtasks state is the same. When rescaling it reassigns states of all tasks to new tasks. If the parallelism is different it assigns the old states to new range of states in a round robin fashion.
> We must make sure that an empty state is not reassigned to running subtasks in {{org.apache.flink.runtime.checkpoint.RoundRobinOperatorStateRepartitioner#repartitionBroadcastState}}
--
This message was sent by Atlassian Jira
(v8.3.4#803005)