DistributedCache when running locally

[Justin Vincent]

Hello,

I am trying to package some config data out to my mappers. Was just testing
while running locally, and I can't get anything to work for me.

~/hadoop/hadoop-0.20.204.0/bin/hadoop jar the_jar.jar com.bar.ApplyMappings
 -files data/config_stuff.txt#config_stuff.txt input_dir output_dir

configure function:

        @Override
        public void configure(JobConf conf) {
            try {
                Scanner s = new Scanner(new File("config_stuff.txt"));
                System.out.println(s.nextLine());
            } catch (Exception e) {
                throw new RuntimeException(e);
            }
        }

I thought distributed cache was supposed to be symlinking the file to my
working directory. To test, I also tried the raw path:

        @Override
        public void configure(JobConf conf) {
            try {
                Scanner s = new Scanner(new
File(DistributedCache.getLocalCacheFiles(conf)[0].toString()));
                System.out.println(s.nextLine());
            } catch (Exception e) {
                throw new RuntimeException(e);
            }
        }

This gives

java.io.FileNotFoundException:
/tmp/hadoop-jvincent/mapred/local/archive/-3183172400467095803_-988649232_1047835448/file/Users/jvincent/projects/foo_project/data/config_stuff.txt
(No such file or directory)

So the file is making it to the local "distributed cache", but not actually
getting copied anywhere useful. Is there something else I need to be
setting up in my JobConf? Or any other ideas? If I pass bogus paths to the
-files flag, I get an exception so it's finding the files and all that
initially.

Thanks!

Justin