[
https://issues.apache.org/jira/browse/HDFS-4127?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13486345#comment-13486345
]
Jing Zhao commented on HDFS-4127:
---------------------------------
Hi Junping, so after you make data node 0&1 not qualified for choosing, I think
you may also need to reset these two nodes back to healthy state at the end of
the new testcase, otherwise the two unqualified nodes will affect the following
testcases.
Another thing is, when calculating the number of replicas still in need for the
log output, do we also need to consider the original size of the results? For
example, when calling the chooseTarget, if there are already 3 nodes chosen
(i.e., we want to increase the number of replicas from 3 to N), after selecting
another S nodes, totalReplicasExpected should be N-3, and the size of the
results is 3+S, and we should expect (N-3)-(3+S)+3=N-3-S more nodes.
> Log message is not correct in case of short of replica
> ------------------------------------------------------
>
> Key: HDFS-4127
> URL: https://issues.apache.org/jira/browse/HDFS-4127
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: name-node
> Affects Versions: 1.0.4, 2.0.2-alpha
> Reporter: Junping Du
> Assignee: Junping Du
> Priority: Minor
> Attachments: HDFS-4127.patch
>
>
> For some reason that block cannot be placed with enough replica (like no
> enough available data nodes), it will throw a warning with wrong number of
> replica in short.
--
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
