[
https://issues.apache.org/jira/browse/HDFS-4606?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13611128#comment-13611128
]
Sanjay Radia commented on HDFS-4606:
------------------------------------
3 variations
# If 3 good replicas: Move replica from a good location (prefer rack local) -
note that there is a move operation that was added for HDFS balancer.
# If 2 good replicas + one stale (stale means DN has missed its heartbeats but
has not been declared dead yet): Copy from one good replicas and NN would
delete the stale one (Need to check if NN currently takes staleness into
consideration for deleting excess replicas).
This case will be common for HBase failover because a region typically fails
over in a minute and NN will not declare a DN as dead till its has missed
heartbeats for longer).
# If 2 good replicas: Copy from one of the good ones - note NN may already be
in the process of replicating. Concern that if we end up with 4 replicas, the
wrong one may be deleted.
> HDFS API to move file replicas to caller's location
> ---------------------------------------------------
>
> Key: HDFS-4606
> URL: https://issues.apache.org/jira/browse/HDFS-4606
> Project: Hadoop HDFS
> Issue Type: New Feature
> Reporter: Sanjay Radia
> Assignee: Sanjay Radia
>
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
