[@stat1007:/home/otto] $ which hdfs-rsync
/usr/local/bin/hdfs-rsync
[@stat1007:/home/otto] $ hdfs-rsync --help
Picked up JAVA_TOOL_OPTIONS: -Dfile.encoding=UTF-8
hdfs-rsync Provides an rsync equivalent to copy data between HDFS and local FS.
Usage: hdfs-rsync [options] src...[dst]


To prevent mistakes between local and remote filesystems, src and dst have to be provided as
fully qualified absolute URIs, for instance file:/home or hdfs:///user/hive.

src should be globs leading to some existing files and dst should be a folder, existing
or at a path where it can be created.

Note: a trailing slash in src (as in /example/src/) is changed to a pattern matching only the
      directory content (as in /example/src/*) mimicking standard rsync behavior.

Syntax for filter/include/exclude rules is similar to the standard rsync one:
 * One rule per command-line option.
 * Order matters as inclusion/exclusion is done picking the first matching pattern.
 * Include/exclude options expect patterns only (use filter if you need modifiers)
 * Extraneous files to be deleted on dst follow include/exclude rules. Use --delete-excluded
   to NOT use the rules on those files.
 * the filter rule format is: RULE[MODIFIERS] PATTERN
     - RULE is either '+' (include) or '-' (exclude).
     - MODIFIERS are optional and can be '!' (negative pattern match) or '/' (match pattern
       against full path even if no '/' or ''), or both.
     - PATTERN is the pattern to match.
   Note: A single space is expected between the rule and modifier char sequence and the pattern.
   Note: Use quotes around the rules when you use wildcard patterns to prevent the shell
         interpreting them.

 About patterns:
 * Pattern wildcard characters are: '*', '**', '?', (see rsync doc). You can escape wildcards
   using '\'. We don't reproduce the character-class wildcards as their definition is not
   present in documentation.
 * Patterns starting with a '/' are anchored, meaning they match only from the root of the
   transfer  (similar to '^' in regex).
 * Patterns with a trailing '/' match only directories.
 * Patterns containing '/' (not counting trailing '/') or '**' are matched against the full
   path (including leading directories). Otherwise it is matched only against the final
   component of the filename.

  --help                Prints this usage text
  --dry-run             Only log instead of actually taking actions (default: false)
  -v, --verbose         Add verbosity to logging (DEBUG)
  -q, --quiet           Remove verbosity from logging (WARN)
  -r, --recursive       Recurse into directories (default: false)
  -p, --perms           Preserve permissions (default: false)
  -t, --times           Preserve modification times (default: false)
  --times-diff <value>  Milliseconds by which modificationTimes can differ and still be considered equal (default: 1000)
  --size-only           Skip files that match in size (default: false)
  -I, --ignore-times    Don't skip files that match size and time (default: false)
  --delete              Delete extraneous files from dst dirs (default: false)
  --delete-excluded     Delete extraneous files from dst dirs even if present in exclude rule (default: false)
  --chmod <value>       affect file and/or directory permissions
  --filter <value>      Add a filter rule
  --include <value>     Add inclusion pattern (this is an alias for: --filter '+ PATTERN')
  --exclude <value>     Add exclusion pattern (this is an alias for: --filter '- PATTERN')
  src...[dst]           Fully qualified URI, one or more sources followed by zero or one destination