[
https://issues.apache.org/jira/browse/HDFS-7759?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14325105#comment-14325105
]
Colin Patrick McCabe commented on HDFS-7759:
--------------------------------------------
This seems like it will lead to a lot of problems. If blocks can't be moved,
then the balancer can't work. It means that if nodes are lost in the cluster,
then we can't re-replicate. Why not simply write a pluggable block placement
policy instead?
> Provide existence-of-a-second-file implementation for pinning blocks on
> Datanode
> --------------------------------------------------------------------------------
>
> Key: HDFS-7759
> URL: https://issues.apache.org/jira/browse/HDFS-7759
> Project: Hadoop HDFS
> Issue Type: Improvement
> Components: datanode
> Reporter: zhaoyunjiong
> Assignee: zhaoyunjiong
> Attachments: HDFS-7759.patch
>
>
> Provide existence-of-a-second-file implementation for pinning blocks on
> Datanode and let admin choosing the mechanism(use sticky bit or
> existence-of-a-second-file) to pinning blocks on favored Datanode.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
