[
https://issues.apache.org/jira/browse/HADOOP-4044?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12638608#action_12638608
]
Doug Cutting commented on HADOOP-4044:
--------------------------------------
Perhaps we're prematurely optimizing here. Perhaps the extra RPCs to resolve
paths won't hurt things that much. The namenode benchmarks I've seen show that
it handles 50k opens/second. How much slower would that be if we always first
resolved the path. In this naive approach we'd add a method which must always
be called before a file is opened (or renamed, etc.) that would return an
'isResolved' boolean and a path. If the path is resolved then it can be used,
otherwise we must resolve it further. The resolution loop could then be
independent of the FileSystem method, so no ThreadLocal or other mechanism
would be required to maintain this state.
{code}
public class ResolvedPath extends Path {}
public class LinkResolution {
public isResolved() { ...}
public Path getUnresolvedPath();
public ResolvedPath getResolvedPath();
}
public class FileSystem {
...
public abstract LinkResolution resolveLinks(Path p);
public abstract FSInputStream open(ResolvedPath p);
private ResolvedPath getResolvedPath(Path p) throws IOException {
LinkResolution resolution = resolveLinks(p);
int count = 0;
while (!resolution.isResolved()) && count < MAX_LINKS) {
p = resolution.getUnresolvedPath();
resolution p.getFileSystem(conf).resolveLinks(p);
}
return resolution.getResolvedPath();
}
public FSInputStream open(Path p, ...) throws IOException {
return open(getResolvedPath(p), ...);
}
...
}
{code}
No exceptions, no thread locals, no anonymous classes, etc. Should we
benchmark this before we rule it out?
> Create symbolic links in HDFS
> -----------------------------
>
> Key: HADOOP-4044
> URL: https://issues.apache.org/jira/browse/HADOOP-4044
> Project: Hadoop Core
> Issue Type: New Feature
> Components: dfs
> Reporter: dhruba borthakur
> Assignee: dhruba borthakur
> Attachments: symLink1.patch, symLink1.patch, 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.
