[
https://issues.apache.org/jira/browse/HDFS-8077?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14486720#comment-14486720
]
Zhe Zhang commented on HDFS-8077:
---------------------------------
Thanks Jing for clarifying. Should we still make a copy if the GS is the same,
but numBytes are different?
{code}
} else if (storedBlock.getNumBytes() != reported.getNumBytes()) {
return new BlockToMarkCorrupt(new Block(reported), storedBlock,
"block is " + ucState + " and reported length " +
reported.getNumBytes() + " does not match " +
"length in block map " + storedBlock.getNumBytes(),
Reason.SIZE_MISMATCH);
{code}
I actually think it's a safer approach to make a copy. So +1 on the patch
pending a decision on the above.
> Erasure coding: fix bug in EC zone and symlinks
> -----------------------------------------------
>
> Key: HDFS-8077
> URL: https://issues.apache.org/jira/browse/HDFS-8077
> Project: Hadoop HDFS
> Issue Type: Sub-task
> Reporter: Zhe Zhang
> Assignee: Jing Zhao
> Attachments: HDFS-8077-000.patch, HDFS-8077.001.patch
>
>
> EC zone manager tries to get XAttr of an inode to determine the EC policy,
> which doesn't work with symlinks. This patch has a simple fix to get rid of
> test failures.
> Ideally we should also add logic to disallow creating symlinks in several
> EC-related scenarios. But since symlinks are disabled in branch-2 and will
> likely be disabled in trunk, this step is skipped now.
> The patch also fixes a small test error around {{getBlockReplication}}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
