[
https://issues.apache.org/jira/browse/HADOOP-3661?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12614927#action_12614927
]
Zheng Shao commented on HADOOP-3661:
------------------------------------
1.
<code>
if (dfs->nowrites || hdfsDelete(dfs->fs, path)) {
syslog(LOG_ERR,"ERROR: hdfs trying to delete the directory %s",path);
return -EIO;
</code>
Can we replace that with "ERROR: hdfs is configured as read-only, cannot delete
the directory %s"
2. Is there any way NOT to hard code "/lib/amd64/server"?
Other things look good to me.
> Normalize fuse-dfs handling of moving things to trash wrt the way hadoop dfs
> does it (only when non posix trash flag is enabled in compile)
> -------------------------------------------------------------------------------------------------------------------------------------------
>
> Key: HADOOP-3661
> URL: https://issues.apache.org/jira/browse/HADOOP-3661
> Project: Hadoop Core
> Issue Type: Improvement
> Components: contrib/fuse-dfs
> Reporter: Pete Wyckoff
> Assignee: Pete Wyckoff
> Attachments: patch0.txt, patch0.txt
>
>
> Currently fuse-dfs is very crude about moving things to Trash. It can only
> move them once since it doesn't use any numbering scheme to prevent
> overwrites.
> Should look at the code that hadoop dfs uses to move things to the trash and
> have the same behavior, i think.
> Leaving it as major because it's pretty annoying and you can't tell what's
> wrong when you get an EIO when deleting something.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
