README

= Amazon Elastic Map Reduce Ruby Client Readme 

Location of the Ruby Client: 

  http://elasticmapreduce.s3.amazonaws.com/elastic-mapreduce-ruby.zip

Welcome the Amazon Elastic MapReduce Ruby client. This package
contains a pure Ruby client for the Amazon Elastic MapReduce Web
Service. Amazon Elastic MapReduce is a service that makes it easy for
researchers, data analysts, and developers to efficiently and
cost-effectively process vast amounts of data using the Amazon EC2 service.
The service launches the EC2 instances on your behalf to
process data, monitor execution and, when the processing is over,
shut down EC2 instances.

It is assumed that the reader is already familiar with Hadoop and the
creation of MapReduce jobs. Additional information on how to write
Map Reduce jobs can be found at the Hadoop website
(http://hadoop.apache.org/).

The version of Hadoop used by Amazon Elastic MapReduce is 0.20.
We also support Hadoop version 0.18.

== Conventions Used in this Document 

Commands that you execute from a bash shell are indented and begin
with $.

  $ echo "This is an example"
  This is an example

The output of the command is sometimes shown directly below the
command.

Sometimes examples contain a value that you should modify for your own
situation, for example:

  $ elastic-mapreduce --jobflow j-ABABABABAB --terminate

In this example, you should replace j-ABABABABAB with the id of the
jobflow that you wish to terminate. Another common example is
s3://mybucket/input. You should replace this with a path in to S3
that contains the data that you want to process.

Other times a value for replacement will be enclosed in angle
brackets. For example, <insert_your_aws_access_id_here>. You should
replace this with your AWS access id including the angle brackets.
For example:

  "access-id": "<insert your AWS access id here>",

becomes

  "access-id": "AAAAAJABASBASBJASAAA",

== Installation and Dependencies

The Amazon Elastic MapReduce Ruby client requires Ruby version 1.8.
It has been tested on Linux computers. The Ruby client can be run on Windows computers,  but has not been fully tested.

To run the client you will need to have signed up for Amazon Elastic
MapReduce Service at http://aws.amazon.com. Amazon Elastic MapReduce
uses Amazon Elastic Compute Cloud (EC2) to run your job flows and Amazon
Simple Storage Service (S3) to store and access your data.  After
completing the sign-up process, you can use Amazon
Elastic Compute Cloud and Simple Storage Service.

=== Step 1: Download and Install Ruby

If you already have Ruby installed, you can skip this step. On Ubuntu
and Debian computers you install Ruby with:

  $ apt-get install ruby1.8

On Redhat computers you can install Ruby with

  yum install ruby

On Macintosh computersyou should already have Ruby installed.

On Windows computers you can install Ruby using the 1-click Windows installer
for Ruby. We recommend you install the final release available at:

  http://rubyinstaller.rubyforge.org/wiki/wiki.pl

You can verify that Ruby is installed by typing

  ruby -v

at the command prompt.

=== Step 2: Download the Amazon Elastic MapReduce Ruby Client

The latest version of the Ruby client can be obtained by

  $ mkdir elastic-mapreduce-ruby
  $ cd elastic-mapreduce-ruby
  $ wget http://elasticmapreduce.s3.amazonaws.com/elastic-mapreduce-ruby.zip

Unzip the archive

  $ unzip elastic-mapreduce-ruby.zip

=== Step 2a: Edit Your Path

If you're running bash or zsh as your shell, then you can add the directory
where you installed the elastic-mapreduce program to your path with the command:

   $ export PATH=$PATH:<directory_where_you_unzipped_elastic_mapreduce_client>

For csh or tcsh

   $ set path = ($path <directory_where_you_unzipped_elastic_mapreduce_client>

=== Step 3: Create a Credentials File

Ensure you are already in the elastic-mapreduce-ruby
directory. Use your AWS access key and private key in the
following command. These credentials are available on the
http://aws.amazon.com website under "Your Account/Access Identifiers"
(top right).

  $ cat > credentials.json
  {
    "access-id": "<insert your AWS access id here>",
    "private-key": "<insert your AWS secret access key here>",
    "key-pair": "<insert the name of your Amazon ec2 key-pair here>",
    "key-pair-file": "<insert the path to the .pem file for your Amazon ec2 key pair here>",
    "region": "<The region where you wish to launch your job flows. Should be one of us-east-1, us-west-1, us-west-2, eu-west-1, ap-southeast-1, or ap-northeast-1, sa-east-1>"    
  }

Windows Users: If you are running a Windows computer then create a
credentials.json file using notepad.exe with the content above inserted between 
the braces.

You do not have to include a key-pair in the credentials file, but it
is a good idea to include it so that when you run job flows you will
be able to log onto the master node to see the log files.

If you don't have an EC2 key-pair, you can create one at:

  https://console.aws.amazon.com/ec2/home#c=EC2&s=KeyPairs

Save the pem file somewhere safe for use later. You will need it to log onto
the master node running your job flow.

If region is not specified, the client will default to us-east-1. Because you 
pay for cross-region data transfer, the region you create your job flows 
should be the one where your S3 input data exists. 

Note: If you have credentials in a file other than credentials.json which is located 
in the current directory, then you can specify a credentials file on the command
line with the following command:

  $ elastic-mapreduce -c <yourcredentialsfile>.json --list

Windows Users: Windows users must specify "ruby" on the command line
instead of "./". The command would be:

  C:\> ruby elastic-mapreduce -c <yourcredentialsfile>.json --list

You can also specify and AWS access id and private key and key pair on the command line.

  $ elastic-mapreduce -a <access-id> -k <private-key> --key-pair <key-pair> --list

Similarly, you can specify a region on the command line.

  $ elastic-mapreduce --region us-east-1 --list

Note that job flow listings are region specific. If you create a job 
flow with region eu-west-1 then you will not be able to see it in the list if you
specify us-east-1. 

== Basic Usage

You can get a summary of supported command line options with:

  $ elastic-mapreduce --help                             

=== Listing Job Flows

The web service supports the following operations: --list, --describe, 
--create, --terminate, --stream, and --jar. To list job flows created in the
last 2 days:

  $ elastic-mapreduce --list
  j-1YE2DN7RXJBWU     FAILED                                                      Example Job Flow
     CANCELLED      Custom Jar                    
  j-3GJ4FRRNKGY97     COMPLETED      ec2-67-202-3-73.compute-1.amazonaws.com      Example job flow
  j-5XXFIQS8PFNW      COMPLETED      ec2-67-202-51-30.compute-1.amazonaws.com     demo 3/24 s1
     COMPLETED      Custom Jar                    

If you have not created any job flows in the last two days 
no output returns from the command. 

The example above shows three job flows created in the last two days. The
indented lines are job flow steps. The columns for a job flow line are
Job Flow Id, Job Flow State, Master Node DNS Name, and Job Flow
Name. The columns for a job flow step line are Step State, and Step
Name.

To get more information about a specific job flow use --describe and
supply the job flow id with the --jobflow parameters.

  $ elastic-mapreduce --describe --jobflow <job_flow_id>
  {
    "JobFlows": [
      {
        "LogUri": null,
        "Name": "Development Job Flow",
        "ExecutionStatusDetail": {
          "EndDateTime": 1237948135.0,
          "CreationDateTime": 1237947852.0,
          "LastStateChangeReason": null,
          "State": "COMPLETED",
          "StartDateTime": 1237948085.0,
          "ReadyDateTime": 1237948085.0
        },
        "Steps": [],
        "Instances": {
          "Ec2KeyName": null,
          "InstanceCount": 1.0,
          "Placement": {
            "AvailabilityZone": "us-east-1a"
          },
          "KeepJobFlowAliveWhenNoSteps": false,
          "TerminationProtected": false,
          "MasterInstanceType": "m1.small",
          "SlaveInstanceType": "m1.small",
          "MasterPublicDnsName": "ec2-67-202-3-73.compute-1.amazonaws.com",
          "MasterInstanceId": "i-39325750",
          "NormalizedInstanceHours": 6,
          "InstanceCount": 3
        },
        "JobFlowId": "j-3GJ4FRRNKGY97"
      }
    ]
  }

You can also list running and starting jobs with:

  $ elastic-mapreduce --list --active

This will list job flows that are starting, running, or shutting
down. You can also list job flows that are in one of several states
with:

  $ elastic-mapreduce --list --state RUNNING --state TERMINATED

This will list job flows that are either running or terminated.

=== Running a Development Job Flow

When developing steps for a job flow it is handy to keep a job flow
running and to add steps to it. This way if the step fails you can
add another step without having to incur the startup cost of a
job flow.

The following command will start a job flow that will continue running and consuming resources until you terminate it.

  $ elastic-mapreduce --create --alive --log-uri s3://my-example-bucket/logs
  Created job flow j-36U2JMAE73054

By default, this command will launch a job flow running on a single m1.small
instance using Hadoop version 0.20. Later, when you have your steps 
running correctly on small set of sample data, you will want to launch job flows 
running on more instance. You can specify the number of instance and 
the type of instance to run with the --num-instances and --instance-type 
options.

The --alive option tells the job flow to keep running even when it has
finished all its steps. The log-uri specifies a location in Amazon S3 for
the log files from your job flow to be pushed. 

--alive option can be safely omitted if you have not yet created a bucket in Amazon
S3. Log files are not pushed to Amazon S3 until 5 minutes after the
step is complete. 

For debugging sessions, you will likely log onto
the master node of your job flow. Specifying a log-uri is required if
you want to be able to read log files from Amazon S3 after the job flow has
terminated.

You can use Hadoop version 0.18 by specifying the --hadoop-version option.

  $ elastic-mapreduce --create --alive --log-uri s3://my-example-bucket/logs \
    --hadoop-version "0.18"

Now that your job flow is created, you can add a streaming step
to the job flow.

  $ elastic-mapreduce --jobflow <j-insert your job id here> --stream
  Added steps to <j-insert your job id here>

Default parameters are added to the streaming step. The name is set to
"Example Streaming Step". The action on failure is set to
"CANCEL_AND_WAIT". This means that if the step fails then subsequent
steps will be cancelled and the job flow will wait for additional
steps to be added before proceeding. The streaming task is a word count
example written in Python and reads input from:

  s3://elasticmapreduce/samples/wordcount/input

You can list the job flow to see the default parameters that have been
set.

  $ elastic-mapreduce --jobflow j-36U2JMAE73054 --describe
  { "JobFlows": [{ 
      "LogUri": null,
      "Name": "Example job flow",
      "ExecutionStatusDetail": { ... },
      "Steps": [
        { "StepConfig": { ... },
          "ExecutionStatusDetail": { ... }
        }
      ],
      "Instances": { ... }
    }]
  }

=== Debugging a Failed Step

First, add a step to the job flow that will fail. If you do not
have a running job flow, look at the section on running job flows above to
see how to start a job flow that will wait for you to add steps.

  $ elastic-mapreduce --jobflow j-36U2JMAE73054 --stream --output hdfs://examples/output
  Added steps to j-36U2JMAE73054

This job will fail because the HDFS path is wrong. It should have
three slashes rather than two after the colon. Now log onto the
master node where the job flow is running to find the log files associated with this step.

If you entered a keypair and a keypair file into the credentials
file you can now log onto the master node with:

  $ elastic-mapreduce --jobflow j-ABABABABABA --ssh 

For this command to work, you are require to have an ssh client installed on
your local computer. Most unix and linux
machines have an ssh client installed. On a Windows computer you will need to install Cygwin, including the Ruby and
openssh-client and execute the elastic-mapreduce client from within
Cygwin.

You can also use PuTTY, but this is significantly more work. See the
tutorial at http://developer.amazonwebservices.com/connect/entry.jspa?externalID=2729.

Once logged into the master node where your job flow is running, you can see log files for your steps in:

  $ ls /mnt/var/log/hadoop/steps/
  1
  $ ls /mnt/var/log/hadoop/steps/1
  controller  stderr  stdout  syslog
  $ cat /mnt/var/log/hadoop/steps/1/syslog
  2009-03-25 18:43:27,145 WARN org.apache.hadoop.mapred.JobClient (main): Use 
    GenericOptionsParser for parsing the arguments. Applications should implement Tool for the same.
  2009-03-25 18:43:28,828 ERROR org.apache.hadoop.streaming.StreamJob (main): Error 
    Launching job : unknown host: examples
  $ exit

The error from Hadoop indicates that it was trying to look for a host
called examples. Look back at our request and see that the output
path was set to hdfs://examples/output. This is incorrect
since we want Hadoop to access the local HDFS system with the path
/examples/output. We need to specify hdfs:///examples/output,
as below.

To fix this, specify the output of the streaming job on the
command line and submit another step to our job flow.

  $ elastic-mapreduce --jobflow j-36U2JMAE73054 --stream --output hdfs:///examples/output

Now list the job flows and watch it complete:

  $ elastic-mapreduce --list -n 5 
  j-36U2JMAE73054     WAITING        ec2-67-202-20-49.compute-1.amazonaws.com     Example job flow
     FAILED         Example Streaming Step        
     COMPLETED      Example Streaming Step        

This time the job succeeded. We can run the job again but this time
output the result to a bucket in Amazon S3. First, create a bucket
in Amazon S3. Note that buckets in Amazon S3 are unique so you will
need to choose a unique name for your bucket.

You can create a bucket using s3cmd which is available on Debian and
Ubuntu systems via apt-get, or using Amazon S3 Organizer which is a
plug-in for Firefox.

If you're not on a Debian or Ubuntu system then s3cmd can be
downloaded from http://freshmeat.net/projects/s3cmd

Amazon S3 Organizer is available at 
https://addons.mozilla.org/en-US/firefox/addon/3247 

To create an Amazon S3 bucket using s3cmd do the following:

  $ s3cmd mb s3://my-example-bucket
  Bucket 's3://my-example-bucket/' created

Note: s3cmd requires s3 paths to be specified using the prefix s3://.
Amazon Elastic MapReduce follows Hadoop’s convention which requires the
prefix s3:// for files in stored in Amazon S3.

Add a step to the job flow to put results into the bucket
you created on S3:

  $ elastic-mapreduce -j j-36U2JMAE73054 --stream \
      --output s3://my-example-bucket/output/1
  Added steps to j-36U2JMAE73054

Note that the protocol of the output URL is s3. This tells Hadoop to
use the S3 Native File System for the output location. The 'host' part
of the URL is the bucket and this is followed by path.

Once you've finished with your job flow, don't forget to terminate it with:

  $ elastic-mapreduce --jobflow j-36U2JMAE73054 --terminate 

And see that it is shutting down with:

  $ elastic-mapreduce --list -n 5

There are other options that you can specify when creating and adding
steps to job flows. Use the "--help" option to list them.

You can also list the log files from the last step in your jobflow
using the --logs argument.

  $ elastic-mapreduce --jobflow j-ABABABABABA --logs

This command requires that you are running on a unix-like
computer with access to an ssh client because it fetches the logs from
the job flow. It also requires that the job flow is still running. If
the job flow has shutdown then the client may hang while attempting to
connect to the master node because the master node has already been
shut down.

==== Enabling Hadoop Debugging

The Amazon Elastic MapReduce tab in the AWS Management Console
(http://console.aws.amazon.com/elasticmapreduce/home) has a debugging
feature that provides access to Hadoop jobs, tasks, and task attempts
as well as log files for steps and task attempts.

Note: To use this feature you must be signed up for SimpleDB. You can sign
up your account for Simple DB at http://aws.amazon.com/simpledb/.

Note: When enabling debugging, you must always specify a log-uri
either in your credentials file, on the command line, or as an
environment variable.

To enable Hadoop Debugging on a job flow run your job flow with the
--enable-debugging option.

  $ elastic-mapreduce --create \
      --name "$USER's Flow with Debugging Enabled" --alive \
      --log-uri s3://mybucket/logs
      --enable-debugging
  Created jobflow j-ABABABABA

Make sure that you sepcified a log-uri on the command line as above or
in your credentials file. This command install a component on your job
flow that pushes information from Hadoop into Amazon Simple DB.

Next add a job flow step to your job flow

  $ elastic-mapreduce --jobflow j-ABABABABA \
    --stream --output hdfs:///output/1

Now you can watch the progress of our job flow in Amazon Elastic
MapReduce tab of the AWS Management Console, by first selecting the
job flow (click refresh if your job flow has not appeared) and then
clicking the "Debug" button.

==== Adding a JAR Step

To add a JAR step you should already have started a job flow. If you have not, see
the "Running a Job Flow" section in this document. 

First start a development job flow

  $ elastic-mapreduce --list --active
  j-36U2JMAE73054     WAITING        ec2-67-202-20-49.compute-1.amazonaws.com     Example job flow
    FAILED         Example Streaming Step        
    COMPLETED      Example Streaming Step        
    COMPLETED      Example Streaming Step        

You can add a JAR step to your job flow with:

  $ elastic-mapreduce --job flow j-36U2JMAE73054 \
    --jar s3://elasticmapreduce/samples/cloudburst/cloudburst.jar \
    --arg s3://elasticmapreduce/samples/cloudburst/input/s_suis.br \
    --arg s3://elasticmapreduce/samples/cloudburst/input/100k.br \
    --arg hdfs:///cloudburst/output/1 \
    --arg 36 --arg 3 --arg 0 --arg 1 --arg 240 --arg 48 --arg 24 \
    --arg 24 --arg 128 --arg 16

Windows Users: The Windows command-line interface does not allow multi-line
commands using the “\” character. You will have to edit these multi-line
examples in notepad to remove the “\” characters and the line breaks.

This will run an example job flow step that downloads and runs the
JAR file. The arguments are passed to the main function in the JAR file.

If your JAR file doesn't have a manifest.mf specifying the main class you will
need to specify the main class on the command line as:

  $ elastic-mapreduce -j j-36U2JMAE73054 \
    --jar        s3://my-example-bucket/wordcount.jar \
    --main-class org.myorg.WordCount \
    --arg        s3://elasticmapreduce/samples/wordcount/input/ \
    --arg        hdfs:///wordcount/output/1

If your job fails then log onto the master node as explained in the
section "Running a JobFlow" and look at the log files to find out why.

==== Adding a JobFlow from JSON

The samples directory included in with Elastic Map Reduce Ruby Client
contains several mutli-step job flows that can be run using the --json
command.

The json files contain variables for bucket names etc that you need to
replace with your own bucket.

  $ elastic-mapreduce -j j-36U2JMAE73054 \
    --json   samples/similarity/lastfm_jobflow.json \
    --param  '<bucket>=my-example-bucket'

This will add the job flow steps described in
samples/freebase/code/freebase_jobflow.json with <bucket> replaced by
my-example-bucket.

=== Samples

The samples directory contains the following sample job flows:

* LastFM Example
** Description: Calculation of Artist Similarity using data From LastFM
** URL: samples/similarity/lastfm_jobflow.json
** Parameters
*** <bucket>  : name of the output bucket

* Freebase Example
** Description: Load popular entries from Freebase into Amazon SimpleDB
** URL: samples/freebase/code/freebase_jobflow.json
** Parameters
*** <bucket>  : name of the output bucket

Additionally there are the following articles explaining how to use
Amazon Elastic MapReduce.

* Sample Job Flows
    http://developer.amazonwebservices.com/connect/kbcategory.jspa?categoryID=263

* CloudBurst Sample Job Flow 
    http://developer.amazonwebservices.com/connect/entry.jspa?externalID=2272

* WordCount Sample Job Flow 
    http://developer.amazonwebservices.com/connect/entry.jspa?externalID=2273

* Similarities Sample Job Flow 
    http://developer.amazonwebservices.com/connect/entry.jspa?externalID=2274

* Freebase Sample Job Flow 
    http://developer.amazonwebservices.com/connect/entry.jspa?externalID=2275

=== Running a Custom Jar

This section assumes that you have already run a streaming jar and
that you are comfortable starting and debugging job flows.

==== Compiling and Uploading a Jar

First download a copy of Hadoop 0.18.3 from the Hadoop website. 

  http://hadoop.apache.org/core/

Unpack the tar.gz archive somewhere handy like
/home/name/hadoop-0.18.3 or c:\hadoop if you are running on windows.

Download Eclipse from http://www.eclipse.org/downloads/, choosing the "Eclipse IDE
for Java Developers" version.

In Eclipse, create a new Java Project. In the libraries tab of the
build path settings, add all of the Hadoop JAR files from the directory where
you unpacked Hadoop.

Next, create a main class in your project. An example main class is
available here:

  http://hadoop.apache.org/core/docs/r0.18.3/mapred_tutorial.html. 

You can cut and paste the WordCount v1.0 source code from the browser
into Eclipse.

Now get Eclipse to export a JAR file by selecting "Export" from the File
Menu and then choosing Jar. Select an output location for the JAR file on
your disk and click through the next several pages until you are able to specify a
Main-Class in the Java manifest. Choose the main function you added in
the previous step.


Upload your JAR file to s3. To upload a file, you can use s3cmd which is
available under Debian and Ubuntu. On Windows computers, a good option is S3
Organizer which is a Firefox plugin. For this example, I'm going to use
s3cmd.

You will have to select your own unique bucket name when running this
example as the my-example-bucket already exists.

  $ s3cmd mb s3://my-example-bucket
  Bucket 's3://my-example-bucket/' created

  $ s3cmd put --force build/wordcount.jar s3://my-example-bucket/wordcount.jar

Note that by default s3cmd sets the access control on the created bucket
and uploaded file to private which is probably what is wanted. If you want
to make your step accessible to other people, you will need to set the
access control on the bucket and the file to make them readable.

Now you are ready to execute a jar step on one of your job flows. If you
don't have job flow running, look at the section on running a job flow
to learn how to start a job flow.

Remember to replace your own job flow id and the name
of your own bucket in the jar name and the second --arg parameter.

  $ elastic-mapreduce --jobflow j-C019299B1X \
    --jar s3://my-example-bucket/wordcount.jar \
    --arg s3://elasticmapreduce/samples/wordcount/input 
    --arg s3://my-example-bucket/output

If you didn't specify the main-class in the manifest then you can
specify that now on the command line with the --main-class option.

  $ elastic-mapreduce --jobflow j-C019299B1X \
    --jar        s3://my-example-bucket/wordcount.jar \
    --main-class org.myorg.WordCount \
    --arg        s3://elasticmapreduce/samples/wordcount/input \
    --arg        s3://my-example-bucket/output

You can watch the job flow by listing it.

  $ elastic-mapreduce --list j-C019299B1X

If the step fails then take a look at the section on debugging a
job flow to find out how to log onto the master node and look at the
logs.

=== Running a Pig Program 

Pig provides a high level data manipulation language called Pig
Latin. Amazon Elastic MapReduce supports Pig version 0.3.0 See
http://hadoop.apache.org/pig/ for additional details.

==== Running Pig in Interactive Mode

To run a pig program in interactive mode you need to create a jobflow
that will stay alive until you terminate it.

  $ elastic-mapreduce --create --alive --name "Testing PIG -- $USER" \
      --num-instances 5 --instance-type c1.large \
      --hive-interactive
  Created jobflow j-ABABABABABAB

Next ssh to the master node and run pig

  $ elastic-mapreduce --jobflow j-ABABABABABAB --ssh 
  ...
  ec2-12.13.14 $ pig
  grunt>

You are now running Pig in interactive mode and can execute Pig Latin statements.

==== Running Pig in Batch Mode

To run Pig in batch mode by adding a step to a development job flow you
run:

  $ elastic-mapreduce --jobflow j-ABABABABABA \
     --pig-script --args s3://mybucket/myscript.pig

Or to run a job flow that will execute a Pig script run:

  $ elastic-mapreduce --create \
     --pig-script \
       --args -p,INPUT=s3://mybucket/input,-p,OUTPUT=s3://mybucket/output \
       --args s3://mybucket/myscript.pig

Note that the script argument must come last, after the parameters that
are passed to the script. These parameters become available within the
Pig script being executed as $variable. In the above example, the
script can make reference to $input and $output within Pig string
literals.

You can also execute multiple Pig scripts
within a single job flow as

  $ elastic-mapreduce --create \
     --pig-script --step-name "Script 1" --args s3://mybucket/myscript.pig \
     --pig-script --step-name "Script 2" --args s3://mybucket/mysecondscript.pig

You can also specify multiple scripts to be added to a development
Job flow in the same way.

=== Running a Hive Program

Hive is a high level data processing language quite similar to SQL but
with a number of map reduce extensions. Amazon Elastic MapReduce
supports Hive version 0.5 on Hadoop version 0.20 and Hive version 0.4
on Hadoop version 0.18. For both the versions of Hive, we provide additional 
patches that allow better integration with Amazon S3.

Please note that Hive automatically defaults to 0.5 for Hadoop 0.20 and
0.4 for Hadoop 0.18. Currently Amazon Elastic MapReduce does not support specifying the Hive
version.

Hive 0.5 provides additional features like Percentile function, faster
recovery of partitions from S3, insert query results into an s3 directory
location, and reading lzo, bzip compressed input files.

==== Running Hive in Interactive Mode

To run in interactive mode you need to create a job flow that will stay
alive until you terminate it.

  $ elastic-mapreduce --create --alive --name "Testing Hive -- $USER" \
      --num-instances 5 --instance-type c1.large \
      --hive-interactive
  Created jobflow j-ABABABABABAB

Next ssh to the master node and run hive

  $ elastic-mapreduce --jobflow j-ABABABABABAB --ssh 
  ...
  ec2-12.13.14 $ hive
  hive>

You are now running Hive in interactive mode and execute Hive queries.

==== Running Hive in Batch Mode

To execute a Hive script stored in S3 as a part of a job flow create the
Job flow with a step that executes the Hive script

  $ elastic-mapreduce --create \
     --hive-script --args s3://mybucket/myquery.q \
     --args -d,INPUT=s3://mybucket/input,-d,OUTPUT=s3://mybucket/output

The --args option provides arguments to the Hive script. The first
argument is the location of the script on S3. Next,the -d argument provides a method to pass values into the script.

Within Hive scripts these parameters are available as ${variable}. In the above example ${INPUT} and ${OUTPUT} would be replaced with
the values that were passed in. These variables are substituted as a
pre-processing step and so may occur anywhere within a Hive script.

You can also add a Hive script to a development job flow, for example:

  $ elastic-mapreduce --jobflow j-ABABABABABA \
     --hive-script --args s3://mybucket/myquery.q \
     --args -d,INPUT=s3://mybucket/input,-d,OUTPUT=s3://mybucket/output

This is useful when developing and testing scripts as in the case where
the script fails. You can add a new step to the development job flow
without having to wait for a new job flow to start.

=== Terminating a Job Flow

All job flows that have been created with the --alive option will run,
consuming instance hours until they are terminated.

A job flow can be terminated by specifying the job flow id.

  $ elastic-mapreduce --terminate --jobflow j-C019299B1X

=== Environment Variables

The command line client accepts configuration via environment
variables. The following environment variables are supported:

    ELASTIC_MAPREDUCE_ACCESS_ID
    ELASTIC_MAPREDUCE_PRIVATE_KEY
    ELASTIC_MAPREDUCE_KEY_PAIR
    ELASTIC_MAPREDUCE_KEY_PAIR_FILE
    ELASTIC_MAPREDUCE_LOG_URI
    ELASTIC_MAPREDUCE_REGION
    ELASTIC_MAPREDUCE_ENABLE_DEBUGGING

The environment variables will override settings in the credentials
file, but will be overridden by options passed in on the command line.

=== Bootstrap Actions

A bootstrap action is a script that is run on all nodes of a job flow
prior to Hadoop starting on that node. A job flow will fail if
bootstrap action fails by returning a non-zero exit code on the master
node, or on more than 10% of the slave nodes. A job flow will also
fail if Hadoop fails to start on the master node after the bootstrap
action has run.

For example, to execute a bootstrap action, first upload your script, called 
action.sh, to your bucket in S3 called s3://mybucket

  $ s3cmd put action.sh s3://mybucket/bootstrap-actions/action.sh

Next, start a job flow and specify the bootstrap action:

  $ elastic-mapreduce --create --alive \
     --name "My Boostrap Action" \
     --bootstrap-action s3://mybucket/bootstrap-actions/action.sh \
     --arg first-argument \
     --arg second-argument

Bootstrap actions may only be specified when a job flow is
created. This is different to steps which may be specified for a
running job flow. When creating a job flow however one may combine
several bootstrap actions and job flow steps. The bootstrap actions
will be executed in the order they are presented.

  $ BUCKET=s3://mybucket/
  $ INPUT=$BUCKET/input
  $ OUTPUT=$BUCKET/output
  $ elastic-mapreduce --create \
      --name "My Example Job Flow" \
      --boostrap-action $BUCKET/bootstrap-actions/action.sh \
        --arg hello  \
        --arg world \
      --hive-script \
        --args s3://mybucket/hive-queries/myquery.q \
        --args -d,INPUT=$INPUT,-d,OUTPUT=$OUTPUT \
      --bootstrap-action $BUCKET/bootstrap-actions/action.sh \
        --args hello,world

Note that the bootstrap actions are always executed before steps and
before Hadoop is running on the node.

Log files from bootstrap actions are located on each node in the
directory:

  /mnt/var/log/bootstrap-actions/<action_number>

where <action_number> is the number of the bootstrap action that was
run. For example,. 1 for the first action, 2 for the second etc. The bootstrap
action logs are also pushed to your LogURI if you specified one in
the directory

  s3://<log_uri>/<jobflow_id>/node/<node_id>/bootstrap-actions/<action_number>

If a bootstap action fails, then an error message containing the
failure condition is located in the LastStateChangeReason field
available when using --describe on the job flow that failed.

=== Example Bootstrap Actions

This section contains a list of example bootstrap actions that are
provided by Amazon Elastic MapReduce.

==== Configure Hadoop

The configure-hadoop bootstrap action allows you specify Hadoop site
configuration that must be set before Hadoop starts. It may be called
as:

  $ elastic-mapreduce --create --alive \
      --name "My Example Jobflow" \
      --bootstrap-actions s3://elasticmapreduce/bootstrap-actions/configure-hadoop
        --arg --site-config-file
        --arg s3://mybucket/config.xml
        --arg -s 
        --arg mapred.tasktracker.map.tasks.maximum=2

This will merge setting from the file s3://mybucket/config.xml into
the Hadoop site config file and will additionally set the value of
mapred.tasktracker.map.tasks.maximum to 2 in the Hadoop site config
file.

For more usage information you can download the script and run it
with no arguments. In this case it will output its usage information.

==== Configure Daemons

The configure daemons script allows you to control the amount of
memory allocated to different Hadoop daemons. It may be called as:

  $ elastic-mapreduce --create --alive \
      --name "My Example Jobflow" \
      --bootstrap-actions s3://elasticmapreduce/bootstrap-actions/configure-daemons \
       --arg --namenode-heap-size=2048 \
       --arg --namenode-opts=-XX:GCTimeRatio=19 \

This will set the heap size allocated to the NameNode to be 2048
megabytes and it will set the GCTimeRatio used by the java's garbage
collector to be 19.

For more usage information you can download the script onto an
instance and run it with no arguments. In this case it will output its
usage information.

==== Run-if

The run-if script allows you to run another script conditionally. It may be called as:

   elastic-mapreduce --create --alive \
      --name "My Example Jobflow" \
      --bootstrap-actions s3://elasticmapreduce/bootstrap-actions/run-if
      --args instance.isMaster=true,s3://mybucket/myscript,hello,world

This will run the script located at s3://mybucket/myscript, passing the
arguments "hello" and "world" only on the master node. If myscript
fails then run-if command will fail and so the job flow will fail.

For more usage information, you can download the script onto an
instance and run it with no arguments. In this case it will output its
usage information onto screen.

The first argument is a conditional that reads from one of the JSON
configuration files available on every instance. The files contain
information that may be useful for bootstrap action authors and with
the run-if script. The file locations and their contents are listed
below. When referencing files from the run-if script, you only need
to specify the file name, excluding ".json".

/mnt/var/lib/info/instance.json
 boolean  isMaster
 boolean  isRunningNameNode
 boolean  isRunningDataNode
 boolean  isRunningJobTracker
 boolean  isRunningTaskTracker

/mnt/var/lib/info/job-flow.json
 string   jobFlowId
 long     jobFlowCreationInstance
 integer  instanceCount
 string   masterInstanceId
 string   masterPrivateDnsName
 string   masterInstanceType
 string   slaveInstanceType
 string   hadoopVersion

== Using the Ruby Client as a Library

The program elastic-mapreduce is an example of how to use the
Amazon Elastic MapReduce Ruby library. Most of the library is
devoted to parsing command line arguments and translating them into
web service calls. 

To use the web service directly in your Ruby programs you need:

  $LOAD_PATH << File.dirname(__FILE__)
  require 'amazon/coral/elasticmapreduceclient'

  config = {
    :endpoint            => "https://elasticmapreduce.amazonaws.com",
    :ca_file             => File.join(File.dirname(__FILE__), "cacert.pem"),
    :aws_access_key      => my_access_id,
    :aws_secret_key      => my_secret_key,
    :signature_algorithm => :V2
  }
  client = Amazon::Coral::ElasticMapReduceClient.new_aws_query(config)

  puts client.DescribeJobFlows.inspect
  puts client.DescribeJobFlows('JobFlowId' => 'j-ABAYAS1019012').inspect

You can use the retry delegator to make your client retry if it gets
connection failures.

  $LOAD_PATH << File.dirname(__FILE__)
  require 'amazon/coral/elasticmapreduceclient'
  require 'amazon/retry_delegator'

  config = {
    :endpoint            => "https://elasticmapreduce.amazonaws.com",
    :ca_file             => File.join(File.dirname(__FILE__), "cacert.pem"),
    :aws_access_key      => my_access_id,
    :aws_secret_key      => my_secret_key,
    :signature_algorithm => :V2
  }
  client = Amazon::Coral::ElasticMapReduceClient.new_aws_query(config)

  is_retryable_error_response = Proc.new do |response| 
    if response == nil then
      false
    else
      ret = false
      if response['Error'] then 
        # don't retry on 'Timeout' because the call might have succeeded
        ret ||= ['InternalFailure', 'Throttling', 'ServiceUnavailable'].include?(response['Error']['Code'])
      end
      ret 
    end
  end

  client = Amazon::RetryDelegator.new(client, :retry_if => is_retryable_error_response)

  puts client.DescribeJobFlows.inspect
  puts client.DescribeJobFlows('JobFlowId' => 'j-ABAYAS1019012').inspect

More information about the operations of the Amazon Elastic MapReduce
Web Service is available in the technical documentation found at http://aws.amazon.com/documentation/elasticmapreduce/.