Setting up Hadoop/YARN/Spark/Hive on Mac OSX

[Reposted from my blogger]

If you are like me, who loves to have everything you are developing against working locally in a mini-integration environment, read on

Here, we attempt to get some pretty heavy-weight stuff working locally on your mac, namely

1. Hadoop (Hadoop2/HDFS)
2. YARN (So you can submit MR jobs)
3. Spark (We will illustrate with Spark Shell, but should work on YARN mode as well)
4. Hive (So we can create some tables and play with it)

We will use the latest stable Cloudera distribution, and work off the jars. Most of the methodology is borrowed from here, we just link the four pieces together nicely in this blog.

Download Stuff

First off all, make sure you have Java 7/8 installed, with JAVA_HOME variable setup to point to the correct location. You have to download the CDH tarballs for Hadoop, Zookeeper, Hive from the tarball page (CDH 5.4.x page) and untar them under a folder (refered to as CDH_HOME going forward) as hadoop, zookeeper

$ ls $HOME/bin/cdh/5.4.7
hadoop                          hadoop-2.6.0-cdh5.4.7.tar.gz    hive-1.1.0-cdh5.4.7             hive-1.1.0-cdh5.4.7.tar.gz      zookeeper                       zookeeper-3.4.5-cdh5.4.7.tar.gz

While you are at it, also grab what version of Spark (pre-built for Hadoop 2.6x) from here, and untar to a directory like below, which we will call $SPARK_INSTALL

$ ls $HOME/bin/spark-1.5.0-bin-hadoop2.6/
CHANGES.txt LICENSE     NOTICE      R           README.md   RELEASE     bin         conf        data        ec2         examples    lib         python      sbin

You may also want to setup a bunch of variables early on, to be of use later

CDH="5.4.7"
export HADOOP_HOME="$HOME/bin/cdh/${CDH}/hadoop"
export ZK_HOME="$HOME/bin/cdh/${CDH}/zookeeper"
export SPARK_INSTALL="$HOME/bin/spark-1.5.0-bin-hadoop2.6"
export PATH=${JAVA_HOME}/bin:${HADOOP_HOME}/bin:${HADOOP_HOME}/sbin:${ZK_HOME}/bin:${SPARK_INSTALL}/bin:${PATH}

Tip 1: If you are using jenv to manage your versions, then you might need the following additional lines in your .bashrc/.bash_profile.

eval "$(jenv init -)"
export JAVA_HOME="$HOME/.jenv/versions/`jenv version-name`"
alias jenv_set_java_home='export JAVA_HOME="$HOME/.jenv/versions/`jenv version-name`"'

Tip 2: Don’t accidentally, name your Spark install dir, SPARK_HOME, Hive does things with it, which you may not like.

Setup Hadoop/YARN

The page we pointed to before, is an excellent resource for doing this already, I will just point out some additional configs I had to add, as I brought in Hive, to make things easier to debug

To etc/hadoop/core-site.xml (to let Hive queries impersonate)

<property>
      <name>hadoop.proxyuser.root.hosts</name>
      <value>*</value>
 </property>
 <property>
      <name>hadoop.proxyuser.root.groups</name>
      <value>*</value>
</property>

To etc/hadoop/yarn-site.xml (to let Hive queries leave a debuggable log)

<property>
  <name>yarn.nodemanager.remote-app-log-dir</name>
  <value>hdfs://localhost:8020/tmp/yarn-logs</value>
  <description>Where to aggregate logs to.</description>
</property>
<property>
  <name>yarn.log-aggregation.retain-seconds</name>
  <value>3600</value>
  <description>Number of seconds to retain logs for</description>
</property>

Make sure, you can start HDFS & YARN locally

Setup Hive

Go into the CDH_HOME/hive-1.1.0-cdh5.4.7 folder and follow the quickstart to build Hive. Basically a command like below

mvn clean package -Phadoop-2,dist

Once you are past the basic steps of quickstart, make a hive-site.xml like below and copy to your hadoop install

$ cat $HADOOP_HOME/etc/hadoop/hive-site.xml
<?xml version="1.0" encoding="UTF-8"?>
<configuration>
  <property>
    <name>hive.metastore.uris</name>
    <value>thrift://localhost:10000</value>
  </property>
</configuration>

Once this is done, you should be able to start a metastore server

[apache-hive-1.1.0-cdh5.4.7-bin]$ bin/hive --service metastore -p 10000

Open up a cli (create table & do a small query)

[apache-hive-1.1.0-cdh5.4.7-bin]$ bin/hive --hiveconf hive.metastore.uris=thrift://localhost:10000
readlink: illegal option -- f
usage: readlink [-n] [file ...]

WARNING: Hive CLI is deprecated and migration to Beeline is recommended.
hive> CREATE TABLE pokes (foo INT, bar STRING);
OK
Time taken: 0.651 seconds
hive> select count(*) from pokes;
Query ID = vinoth_20160523115454_527e550c-7318-4ffc-a49f-248ca119c5a8
Total jobs = 1
Launching Job 1 out of 1
Number of reduce tasks determined at compile time: 1
In order to change the average load for a reducer (in bytes):
  set hive.exec.reducers.bytes.per.reducer=<number>
In order to limit the maximum number of reducers:
  set hive.exec.reducers.max=<number>
In order to set a constant number of reducers:
  set mapreduce.job.reduces=<number>
Hadoop job information for Stage-1: number of mappers: 1; number of reducers: 1
2016-05-23 11:54:28,958 Stage-1 map = 0%,  reduce = 0%
2016-05-23 11:54:34,119 Stage-1 map = 100%,  reduce = 0%
2016-05-23 11:54:39,249 Stage-1 map = 100%,  reduce = 100%
Ended Job = job_1464029642280_0001
MapReduce Jobs Launched:
Stage-Stage-1: Map: 1  Reduce: 1   HDFS Read: 6476 HDFS Write: 2 SUCCESS
Total MapReduce CPU Time Spent: 0 msec
OK
0
Time taken: 18.299 seconds, Fetched: 1 row(s)
hive>

Setup Spark

Spark is super simple, just need to point Spark to the Hadoop installation, that has not only the Hadoop configs, but also the Hive config (this is why we cp-ed hive-site.xml before)

$ cd $SPARK_INSTALL
$ export HADOOP_CONF_DIR=$HADOOP_HOME/etc/hadoop
$ spark-shell --driver-class-path $HADOOP_CONF_DIR 
scala> sqlContext.sql("show tables").show()
scala> sqlContext.sql("describe pokes").show()
scala> sqlContext.sql("select count(*) from pokes").show()

Voila!! (not really a quick thing to do, but once you have done it once, then you can setup debugger etc and its all golden)