[
https://issues.apache.org/jira/browse/HADOOP-14676?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16097058#comment-16097058
]
Konstantin Shvachko commented on HADOOP-14676:
----------------------------------------------
I traced it all the way to 2.6.1.
DataNodes on our cluster were using too much system time and it was traced to
frequent {{statfs}}. Thanks [~ga] for digging into this.
The workaround is to actually specify the default value in {{core-site.xml}},
then it will not use the fake default. But we should fix it.
> Wrong default value for "fs.du.interval"
> ----------------------------------------
>
> Key: HADOOP-14676
> URL: https://issues.apache.org/jira/browse/HADOOP-14676
> Project: Hadoop Common
> Issue Type: Bug
> Components: common, conf, fs
> Affects Versions: 2.6.1
> Reporter: Konstantin Shvachko
>
> According to {{core-default.xml}} the default value of {{fs.du.interval = 60
> sec}}. But the implementation of {{DF}} substitutes 3 sec instead. The
> problem is that {{DF}} uses outdated constant {{DF.DF_INTERVAL_DEFAULT}}
> instead of the correct one
> {{CommonConfigurationKeysPublic.FS_DU_INTERVAL_DEFAULT}}.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
