[
https://issues.apache.org/jira/browse/HDFS-7735?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
zhaoyunjiong updated HDFS-7735:
-------------------------------
Attachment: HDFS-7735.patch
This patch try to reduce the impact in two ways:
# Adjust dfs.namenode.decommission.interval from 30 to 300 to reduce the check
frequency
# Remove blocks already have enough live replicas from decommissioning nodes to
reduce check time
> Optimize decommission Datanodes to reduce the impact on NameNode's performance
> ------------------------------------------------------------------------------
>
> Key: HDFS-7735
> URL: https://issues.apache.org/jira/browse/HDFS-7735
> Project: Hadoop HDFS
> Issue Type: Improvement
> Components: namenode
> Reporter: zhaoyunjiong
> Assignee: zhaoyunjiong
> Attachments: HDFS-7735.patch
>
>
> When decommission DataNodes, by default, DecommissionManager will check
> progress every 30 seconds, and it will hold writeLock of Namesystem. It
> significantly impact NameNode's performance.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
