Permalink
Find file
13ba9a2 Mar 22, 2013
@michiard @antoniobarbuzzi @lucab
99 lines (53 sloc) 2.56 KB

HADOOP Cheat Sheet

Working with HDFS from the command line

hadoop dfs <CMD>

Inspect files

  • -ls <path>: list all files in <path>
  • -cat <src>: print <src> on stdout
  • -tail [-f] <file>: output the last part of the <file>
  • -du <path>: show <path> space utilization

Create/remove files

  • -mkdir <path>: create a directory
  • -mv <src> <dst>: move (rename) files
  • -cp <src> <dst>: copy files
  • -rmr <path>: remove files

Copy/Put files from a remote machine into the HADOOP cluster

  • -copyFromLocal <localsrc> <dst>: copy a local file to the HDFS
  • -copyToLocal <src> <localdst>: copy a file on the HDFS to the local disk

HELP

  • -help [cmd]: hopefully this is self-describing

Examples:

hadoop dfs -ls /

hadoop dfs -copyFromLocal myfile remotefile

Launching Hadoop Jobs - Command line

  • Copy the jar file of your job to the client machine (let's call it machine_name)

scp localJarFile studentXX@machine_name:~/

  • SSH to machine_name:

ssh studentXX@machine_name

  • Launch the job:

hadoop jar jarFile.jar ClassNameWithPackage [job args]

Note that if the output directory exists (and you don't want it) you need to remove it:

hadoop dfs -rmr output

Example:

hadoop jar fr.eurecom.dsg.WordCount /user/hadoop/wikismall.xml output 2

Reading (Textual) Input Data in the Mapper

This is the class you're looking for: org.apache.hadoop.mapreduce.lib.input.TextInputFormat<K,V>

Precisely, this is the class hierarchy:

java.lang.Object

org.apache.hadoop.mapreduce.InputFormat<K,V>

org.apache.hadoop.mapreduce.lib.input.FileInputFormat<LongWritable,Text>

org.apache.hadoop.mapreduce.lib.input.TextInputFormat

Basically, this is an InputFormat specifically designed for plain text files. Files are broken into lines. Either linefeed or carriage-return are used to signal end of line. Keys are the position in the file, and values are the line of text. You need to take care of the following:

Key Type: LongWritable

Value Type: Text

Writing (Textual) Output Data in the Reducer

This is the class you're looking for: org.apache.hadoop.mapreduce.lib.output.TextOutputFormat<K,V>

Precisely, this is the class hierarchy:

java.lang.Object

org.apache.hadoop.mapreduce.OutputFormat<K,V>

org.apache.hadoop.mapreduce.lib.output.FileOutputFormat<K,V>

org.apache.hadoop.mapreduce.lib.output.TextOutputFormat<K,V>

Essentially, this OutputFormat writes plain text files. TextOutputFormat calls toString() for each key and value pair in output, so any (Writable) type can be used.