[
https://issues.apache.org/jira/browse/HDFS-245?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12759293#action_12759293
]
Eli Collins commented on HDFS-245:
----------------------------------
*Apologies for the dupe, looks like jira stripped out the quoted text in my
email above.*
Thanks Nigel, Allen, Robert for taking a look. Will update the doc with
feedback.
bq. Loops should be avoided by having the client limit the number of links it
will traverse What about loops within a filesystem? Does the NN also limit the
number of links it will traverse?
Good point, the NN should do the same, use the same limit for consistency.
bq. You give examples of commands the operate on the link and examples that
operate on the link target and examples of those that depend on a trailing
slash. Given this is a design, can you be explicitly and enumerate the
commands for each of there? For instance, setReplication, setTimes, du, etc.
Will do, how about as part of the user guide?
bq. What is the new options for fsck to report dangling links? What does the
output look like?
Agree with Allen. Updated to read "fsck should not follow or report dangling
links"
bq. What is the new option for distcp to follow symlinks?
Probably don't need one, seems like distcp should following links and...
bq. If distcp doesn't follow symlinks, I assume it just copies the symlink. In
this case, is the symlink adjusted to point to the source location on the
source FS?
would not adjust links when it's copying them.
bq. What does the ls output look like for a symlink?
Not sure the design doc needs to specify but how about the usual fileName ->
path?
bq. Do symlinks contribute bytes toward a quota?
Agree w/ Rob, links should not count towards quota (they have no storage after
all).
bq. access control properties of links are ignored, checks are always performed
against the link target (if it resides in an HDFS file system). The ch*
operations should operate directly on the target.
bq. I disagree. We need chown -h (POSIX) and chown -H, chown -L, and chown -P
(SVID?) for the same reasons that those OSes support them.
You're right, the above should be chmod (not ch*). And checks against the link
target should be performed regardless of what file system it resides on. And
yes ch* utilities need to be link-aware.
bq. We also need test -h and test -L supported.
Agreed. Seems like these should all go in the user guide, unless people want
them spelled out in the design doc.
Wrt lsr and rmr, I'd make them not follow links by default and add options that
do (can detect loops by remembering the paths they've visited).
+1 to Doug's idea of exposing links solely via FileContext, so that day zero
behavior ignores links and clients get updated to be link aware as necessary.
Are people OK updating them as they move to FileContext?
Thanks,
Eli
> Create symbolic links in HDFS
> -----------------------------
>
> Key: HDFS-245
> URL: https://issues.apache.org/jira/browse/HDFS-245
> Project: Hadoop HDFS
> Issue Type: New Feature
> Reporter: dhruba borthakur
> Assignee: dhruba borthakur
> Attachments: 4044_20081030spi.java, designdocv1.txt,
> HADOOP-4044-strawman.patch, symlink-0.20.0.patch, symLink1.patch,
> symLink1.patch, symLink11.patch, symLink12.patch, symLink13.patch,
> symLink14.patch, symLink15.txt, symLink15.txt, symlink16-common.patch,
> symlink16-hdfs.patch, symlink16-mr.patch, symlink17-common.txt,
> symlink17-hdfs.txt, symLink4.patch, symLink5.patch, symLink6.patch,
> symLink8.patch, symLink9.patch
>
>
> HDFS should support symbolic links. A symbolic link is a special type of file
> that contains a reference to another file or directory in the form of an
> absolute or relative path and that affects pathname resolution. Programs
> which read or write to files named by a symbolic link will behave as if
> operating directly on the target file. However, archiving utilities can
> handle symbolic links specially and manipulate them directly.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
