[
https://issues.apache.org/jira/browse/HADOOP-12973?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15218747#comment-15218747
]
Elliott Clark commented on HADOOP-12973:
----------------------------------------
bq.Hi Elliott Clark I wonder what was the concern about replacing the call to
DU?
My understanding is that replacing du with df means that we will lose the
ability to tell how large the distributed cache is. So for people who rely on
MR/Yarn a lot then it's possible to get into a bad situation. I haven't really
looked into that too much. My primary use case is around HBase where a du
running and walking all the inodes causes a significant IO spike and latency
outliers.
bq.With regard to the issues described in HDFS-9923, do you think that your
solution also fixes the inconsistent exception handling of DU?
I didn't change the exception handling at all, but it is weird and changing it
to just log a warn seems like a good idea to me.
> make DU pluggable
> -----------------
>
> Key: HADOOP-12973
> URL: https://issues.apache.org/jira/browse/HADOOP-12973
> Project: Hadoop Common
> Issue Type: Sub-task
> Reporter: Elliott Clark
> Assignee: Elliott Clark
> Attachments: HADOOP-12973v0.patch, HADOOP-12973v1.patch,
> HADOOP-12973v2.patch, HADOOP-12973v3.patch, HADOOP-12973v5.patch,
> HADOOP-12973v6.patch
>
>
> If people are concerned about replacing the call to DU. Then an easy first
> step is to make it pluggable. Then it's possible to replace it with something
> while leaving the default alone.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
