[
https://issues.apache.org/jira/browse/HADOOP-4598?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12647709#action_12647709
]
Hairong Kuang commented on HADOOP-4598:
---------------------------------------
UnderReplicatedBlocks.update has an optimization that a under-replication block
does not get added to the queue if a block was under-replicated and its
previous priority is the same as the current priority. This optimization is
based on the assumption that a previously under-replicated block should have
already been in the queue.
For the case that Koji filed, due to for some bugs like HADOOP-4597,
under-replicated blocks did not get to place in the under-replicated block
queue at NameNode startup time. Later on changing a file's replication factor
to be a bigger number does not trigger the move of the file's blocks into the
queue if the replication factor change does not change their priority.
The fix for this problem is to remove the optimization.
> '-setrep' command skips under-replicated blocks
> -----------------------------------------------
>
> Key: HADOOP-4598
> URL: https://issues.apache.org/jira/browse/HADOOP-4598
> Project: Hadoop Core
> Issue Type: Bug
> Components: dfs
> Reporter: Koji Noguchi
> Assignee: Hairong Kuang
> Priority: Minor
>
> When blocks are under-replicated due to some hadoop bugs, like HADOOP-4597,
> calling
> hadoop dfs -setrep <higher_replication> <filename>
> would not schedule the replication for the under-replicated blocks (unless it
> goes to a different priority queue).
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
