Aprendizaje en Grandes Volúmenes de Datos, Clase 17
por Pablo Ariel Duboue, PhD
disponible bajo licencia CC-BY-SA

Presentación:

* feedback
  * alternativa a hadoop que yo recomendaria: message queues. Porque? Mas flexibilidad, funcionan off-line.
  * si se pierden o no andan, llamenme
  * script de antemano: no, es importante que se tomen el tiempo detipearlo. Cut&paste ea demasiado

En máquina:

# seteando una maquina para pseudo-cluster

In the web: http://hadoop.apache.org/docs/current/hadoop-project-dist/hadoop-common/SingleCluster.html#Pseudo-Distributed_Operation

[Following these instructions you can set-up a pseudo-distributed
hadoop in your own machine, in class we logged into a machine --no
longer available-- where the set-up had already taken place, by
executing the code below.]

# note: you need a ssh server running in your machine

PROMPT$ echo 'export JAVA_HOME=/usr/lib/jvm/java-7-openjdk-amd64' >> .profile
PROMPT$ cd hadoop-2.4.1
PROMPT$ echo export HADOOP_PREFIX=$PWD >> ../.profile 
PROMPT$ mv *.java hadoop-2.4.1
PROMPT$ echo 'export HADOOP_CLASSPATH=$JAVA_HOME/lib/tools.jar' >> ~/.profile 
PROMPT$ ./bin/hadoop com.sun.tools.javac.Main WordCount.java
PROMPT$ jar cf wc.jar WordCount*.class
PROMPT$ ssh-keygen -t dsa -P '' -f ~/.ssh/id_dsa
PROMPT$ cat ~/.ssh/id_dsa.pub >> ~/.ssh/authorized_keys
PROMPT$ mv lib/native ./lib/native-out
PROMPT$ emacs etc/hadoop/core-site.xml etc/hadoop/hdfs-site.xml etc/hadoop/hadoop-env.sh

# set etc/hadoop/core-site.xml:
<configuration>
    <property>
        <name>fs.defaultFS</name>
        <value>hdfs://localhost:9000</value>
    </property>
</configuration>

# set etc/hadoop/hdfs-site.xml
<configuration>
    <property>
        <name>dfs.replication</name>
        <value>1</value>
    </property>
</configuration>

# set JAVA_HOME etc/hadoop/hadoop-env.sh

PROMPT$ ./bin/hdfs namenode -format
PROMPT$ ./sbin/start-dfs.sh


[You don't need to ssh if setting up in your own machine.]

# ssh -L 50070:localhost:50070  hadoop@arenero.aprendizajengrande.net
# go to http://localhost:50070/

PROMPT$ ./bin/hdfs dfs -mkdir /user
PROMPT$ ./bin/hdfs dfs -df
PROMPT$ wget https://archive.org/download/2013_common_crawl_index_urls/common_crawl_index_urls.bz2

# 21Gb later... 39.7M/s in 11m 22s

PROMPT$ ./bin/hdfs dfs -mkdir /data
PROMPT$ ./bin/hdfs dfs -put common_crawl_index_urls.bz2 /data/

PROMPT$ ./bin/hadoop jar wc.jar WordCount /data/common_crawl_index_urls.bz2 /user/pablo/outputwc

# too long single mode

PROMPT$ zcat /usr/share/doc/*/README* > README
PROMPT$ ./bin/hdfs dfs -put README /data/
PROMPT$ ./bin/hadoop jar wc.jar WordCount /data/README /user/pablo/outputwc2
PROMPT$ ./bin/hdfs dfs -cat /user/pablo/outputwc2/part-r-00000

# setting YARN

PROMTP$ emacs etc/hadoop/mapred-site.xml etc/hadoop/yarn-site.xml etc/hadoop/yarn-env.sh

# in mapred-site.xml set:
<configuration>
    <property>
        <name>mapreduce.framework.name</name>
        <value>yarn</value>
    </property>
</configuration>

# in yarn-site.xml set:

    <property>
        <name>yarn.nodemanager.aux-services</name>
        <value>mapreduce_shuffle</value>
    </property>

# in yarn-env.sh set JAVA_HOME and HADOOP_YARN_USER

PROMPT$ ./sbin/start-yarn.sh



[You don't need to ssh if setting up in your own machine.]

# ssh -L 50070:localhost:50070 -L 8088:localhost:8088  hadoop@arenero.aprendizajengrande.net

# check http://localhost:8088/