The Gluster Blog

Gluster blog stories provide high-level spotlights on our users all over the world

Hadoop : Gotchas movin to the cluster. Part II


Okay so… in the last post, i tried to build a non-trivial map/r from scratch, and ran it on my machine.   I ran into some issues involving the “glue” that held my map/reduce jobs together.  For example, configuring the classes declared mapper outputs, and properly “telling” hadoop that I didn’t “need” a reducer, etc etc…

Once those were fixed, it ran beautifully on my local machine, generating hundreds of correct mapper outputs, and ending before the reducer stage.

When we ran it in the cluster, we found another interesting error, however.

     So — its the next day, and we tried to run my map/reduce job in the cluster, and it broke immedeately….
Why ? Because I used apache’s commons-io for a few things….So I had to convert the following block of code. Below is a contrast of the original code I used for recursively reading through a text file.  versus the exact same logic, implemented without Commons-io, using the Hadoop FileSystem API.

The important take home lesson for me was that hadoop’s FileSystem API is compatible with both local and distributed file systems, and so it works great for unit tests and standard, local disk io.  However, the additional advantage is that the FileSystem API, by virtue of the Configuration object which it recieves, is modular enough to operate equally effectively on a hadoop cluster, in the hadoop DFS.

The old way  🙁

       /** Old code – only works locally, crashes due to file protocol excption in a cluster ** /

        File rootDirectory = new File(rootPath.makeQualified(FileSystem.get(conf)).toUri());
        //case where we had a file as input.
            return Arrays.asList(new Path(rootDirectory.getAbsolutePath()));
        //Other case, a directory was the input.
        IOFileFilter filter = FileFilterUtils.prefixFileFilter(“part-“);
        IOFileFilter dirFilter = FileFilterUtils.directoryFileFilter();

        Collection<File> c1 = FileUtils.listFiles(rootDirectory, filter, dirFilter);
        List<Path> paths = Lists.newArrayList();
        for(File f1 : c1)
            System.out.println(“\t found json file : “+f1.getAbsolutePath() + ” ” + f1.length()/1000 + ” kb”);
            paths.add(new Path(f1.getAbsolutePath()));
        return paths;

The hadoop way ! 🙂

public static Collection<Path> determineInputs(Path rootPath,Configuration conf) throws IOException
        if(! FileSystem.get(conf).getFileStatus(rootPath).isDir())
            return Arrays.asList(rootPath);
        FileStatus[] c1 = FileSystem.get(conf).listStatus(rootPath);
        List<Path> paths = Lists.newArrayList();
        for(FileStatus f1 : c1)
                System.out.println(“\t found json file : “+f1 + ” ” + f1.getLen()/1000 + ” kb”);
                if(f1.getPath().getName().contains(“part-“) && ! f1.getPath().getName().startsWith(“.”))
        return paths;



  • 06 Dec 2020
    Looking back at 2020 – with g...

    2020 has not been a year we would have been able to predict. With a worldwide pandemic and lives thrown out of gear, as we head into 2021, we are thankful that our community and project continued to receive new developers, users and make small gains. For that and a...

    Read more
  • 27 Apr 2020
    Update from the team

    It has been a while since we provided an update to the Gluster community. Across the world various nations, states and localities have put together sets of guidelines around shelter-in-place and quarantine. We request our community members to stay safe, to care for their loved ones, to continue to be...

    Read more
  • 03 Feb 2020
    Building a longer term focus for Gl...

    The initial rounds of conversation around the planning of content for release 8 has helped the project identify one key thing – the need to stagger out features and enhancements over multiple releases. Thus, while release 8 is unlikely to be feature heavy as previous releases, it will be the...

    Read more