You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@flink.apache.org by "Yun Tang (Jira)" <ji...@apache.org> on 2022/01/17 02:54:00 UTC
[jira] [Assigned] (FLINK-25580) Minor improvements for RocksDBIncrementalCheckpointUtils
[ https://issues.apache.org/jira/browse/FLINK-25580?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Yun Tang reassigned FLINK-25580:
--------------------------------
Assignee: Aitozi
> Minor improvements for RocksDBIncrementalCheckpointUtils
> --------------------------------------------------------
>
> Key: FLINK-25580
> URL: https://issues.apache.org/jira/browse/FLINK-25580
> Project: Flink
> Issue Type: Improvement
> Components: Runtime / State Backends
> Reporter: Aitozi
> Assignee: Aitozi
> Priority: Minor
> Labels: pull-request-available
> Fix For: 1.15.0
>
>
> {{RocksDBIncrementalCheckpointUtils}} use {{intersectGroup.getNumberOfKeyGroups() * overlapFraction * overlapFraction}} to evaluate the score of the state handle. It's meaning is not so explicit. And the rule to choose the handle should be
> 1. Choose the bigger intersect group range first
> 2. If the intersect group range are same, we use the handle with higher overlap fraction, which means less iterator delete operation.
> After we introduce the {{deleteRange}} api, this rule can still work.
--
This message was sent by Atlassian Jira
(v8.20.1#820001)