apache/hadoop · error · PathIsDirectoryException

Is a directory

Error message

Is a directory

What it means

PathIsDirectoryException (message 'Is a directory') thrown by Getmerge.processOptions (CopyCommands.java:83) for the 'hadoop fs -getmerge' command. The destination argument (the last argument) already exists and is a directory. getmerge merges many source files into ONE local output file, so the destination must be a (typically nonexistent) file path, not a directory.

Source

Thrown at hadoop-common-project/hadoop-common/src/main/java/org/apache/hadoop/fs/shell/CopyCommands.java:83

    protected PathData dst = null;
    protected String delimiter = null;
    private boolean skipEmptyFileDelimiter;
    protected List<PathData> srcs = null;

    @Override
    protected void processOptions(LinkedList<String> args) throws IOException {
      try {
        CommandFormat cf = new CommandFormat(2, Integer.MAX_VALUE, "nl",
            "skip-empty-file");
        cf.parse(args);

        delimiter = cf.getOpt("nl") ? "\n" : null;
        skipEmptyFileDelimiter = cf.getOpt("skip-empty-file");

        dst = new PathData(new URI(args.removeLast()), getConf());
        if (dst.exists && dst.stat.isDirectory()) {
          throw new PathIsDirectoryException(dst.toString());
        }
        srcs = new LinkedList<PathData>();
      } catch (URISyntaxException e) {
        throw new IOException("unexpected URISyntaxException", e);
      }
    }

    @Override
    protected void processArguments(LinkedList<PathData> items)
    throws IOException {
      super.processArguments(items);
      if (exitCode != 0) { // check for error collecting paths
        return;
      }
      FSDataOutputStream out = dst.fs.create(dst.path);
      try {
        for (PathData src : srcs) {
          if (src.stat.getLen() != 0) {

View on GitHub (pinned to 2add963021)

Solutions

1. Point the destination at a file name that does not exist: 'hadoop fs -getmerge /hdfs/src/dir /local/merged.txt'
2. Or keep the directory and append a file name to it: 'hadoop fs -getmerge /hdfs/src/dir /local/output/merged.txt'
3. If the directory was created by mistake in a prior step, remove or skip that mkdir
4. Use the -nl option only after the destination is a valid file path (options are unrelated but often appear in the same failing scripts)

Example fix

# before
mkdir -p /local/out
hadoop fs -getmerge /src/dir /local/out      # Is a directory

# after
mkdir -p /local/out
hadoop fs -getmerge /src/dir /local/out/merged.txt

PathData dst = new PathData(new URI(dstArg), getConf());
if (dst.exists && dst.stat.isDirectory()) {
  throw new PathIsDirectoryException(dst.toString());
}

static boolean isValidGetmergeDestination(PathData dst) throws IOException {
  return !(dst.exists && dst.stat.isDirectory());
}

try {
  // run getmerge
} catch (PathIsDirectoryException e) {
  // append an output file name to the directory and retry
}

When it happens

Trigger: 'hadoop fs -getmerge /hdfs/src/dir /local/output' where /local/output is an existing directory; scripts that pass a target directory expecting cp/mv-like semantics; a destination created beforehand with mkdir by the same script.

Common situations: Copy/paste of a 'cp -r' idiom where the last arg is a directory; automation that derives the destination from a directory variable; users expecting getmerge to write merged.txt into the given folder automatically.

Related errors

• Is a directory
• Is a directory
• Is a directory
• Illegal option {}
• Not enough arguments: expected {} but got {}

AI-assisted analysis of apache/hadoop@2add963021 (2026-08-22). Data as JSON: /api/errors/dabbc1733fe99495. Report an issue: GitHub.