[
https://issues.apache.org/jira/browse/FLINK-28701?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Weijie Guo updated FLINK-28701:
-------------------------------
Parent: (was: FLINK-27862)
Issue Type: Bug (was: Sub-task)
> Minimize the explosion range during failover for hybrid shuffle
> ---------------------------------------------------------------
>
> Key: FLINK-28701
> URL: https://issues.apache.org/jira/browse/FLINK-28701
> Project: Flink
> Issue Type: Bug
> Components: Runtime / Coordination
> Affects Versions: 1.16.0
> Reporter: Weijie Guo
> Priority: Major
>
> In hybrid shuffle mode, there are currently two strategies to control
> spilling. For the full spilling strategy, the data is guaranteed to be
> persistent to the disk after task finished. When a failover occurs, if the
> upstream has been finished, the data should be recovered directly from the
> disk file without re-compute. For selective spilling strategy, the entire
> topology must be restarted.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
