[
https://issues.apache.org/jira/browse/IOTDB-5589?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17735528#comment-17735528
]
周沛辰 commented on IOTDB-5589:
----------------------------
# 控制合并调度（选文件）的线程数量
# 修复删除存储组后，合并线程泄漏的bug
> [ compaction ] The "IoTDB-Compaction_Schedule" thread count needs to be
> controlled
> ------------------------------------------------------------------------------------
>
> Key: IOTDB-5589
> URL: https://issues.apache.org/jira/browse/IOTDB-5589
> Project: Apache IoTDB
> Issue Type: Improvement
> Components: mpp-cluster
> Reporter: 刘珍
> Assignee: 周沛辰
> Priority: Major
> Attachments: comp1.conf, comp2.conf, dn_stack.out
>
>
> master_a3869ec_opt
> Each time partition opens a "IoTDB-Compaction_Schedule" thread，it's not a
> good design.
> 示例：
> 在192.168.10.68 （72CPU) ,
> start-standalone.sh启动单节点集群，运行附件中的脚本，生成72个dataregion，每个dataregion
> 下目前只有1个时间分区，所以当前"IoTDB-Compaction_Schedule"线程数有72个，随着时间分区的增加，这个线程也会增大，需要控制，它的主要作用是用来选择合并文件，是比较快的操作，没必要开这么多。
> 纠正：是每个dataregion 有1个 "IoTDB-Compaction_Schedule"
> thread，是否解决，可以讨论。如果dataregion多，是否有必要开这么多选文件的线程。
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
