Vagrant project to spin up a single virtual machine running:
- Hadoop 2.7.3
- Hive 1.2.2
- Spark 2.1.1
- mysql 5.7 (for hive metastore)
- Tez 0.8.5
- Sqoop 1.4.6
- Pig 0.17.0
- flume 1.7.0
- Zeppelin 0.7.2 (with Spark/scala, md, file and JDBC interpreters)
The virtual machine will be running the following services:
- HDFS NameNode + DataNode
- YARN ResourceManager/NodeManager + JobHistoryServer + ProxyServer
- Hive metastore and server2
- Spark history server
- mysql
- zeppelin notebook server (not started by default - see below)
- Download and install VirtualBox
- Download and install Vagrant.
- Go to releases and download and extract the latest source of this project.
- In your terminal change your directory into the project directory (i.e.
cd vagrant-hadoop-spark-hive-<version>
). - Run
vagrant up
to create the VM (NOTE This will take a while the first time as many dependancies are downloaded - subsequent deployments will be quicker as dependancies are cached in theresources
directory). - Execute
vagrant ssh
to login to the VM.
By default map reduce jobs will be executed via Tez to change this to standard MR, change the following parameter in $HADOOP_CONF/mapred-site.xml from: -
<property>
<name>mapreduce.framework.name</name>
<value>yarn-tez</value>
</property>
to
<property>
<name>mapreduce.framework.name</name>
<value>yarn</value>
</property>
Here are some useful links to navigate to various UI's:
- YARN resource manager: (http://10.211.55.101:8088)
- Job history: (http://10.211.55.101:19888/jobhistory/)
- HDFS: (http://10.211.55.101:50070/dfshealth.html)
- Spark history server: (http://10.211.55.101:18080)
- Spark context UI (if a Spark context is running): (http://10.211.55.101:4040)
- Zeppelin notebook (if notebook server is started) (http://10.211.55.101:8080)
Apache Zeppelin together with spark, md, file and JDBC interpreters is installed by default but not started. To manually start the zeppelin daemon run: -
vagrant ssh
zeppelin-daemon.sh start
as ubuntu
user from the command line.
Notebook server can then be accessed via http://10.211.55.101:8080
.
mysql database connection
[email protected]:3306
(for root DBA access - password is 'root')[email protected]:3306
(for hive metastore user - password is 'hive')
Vagrant automatically mounts the folder containing the Vagrant file from the host machine into
the guest machine as /vagrant
inside the guest.
To test out the virtual machine setup, and for examples of how to run MapReduce, Hive and Spark, head on over to VALIDATING.md.
To stop the VM and preserve all setup/data within the VM: -
vagrant halt
or
vagrant suspend
Issue a vagrant up
command again to restart the VM from where you left off.
To completely wipe the VM so that vagrant up
command gives you a fresh machine: -
vagrant destroy
Then issue vagrant up
command as usual.
$ vagrant ssh
$ sudo -sE
$ /vagrant/scripts/stop-spark.sh
$ /vagrant/scripts/stop-hadoop.sh
$ systemctl stop mysql.service
Spark in particular needs quite a bit of memory to run - to work around this a swapspace
daemon is also configured and
started that uses normal disk to dynamically allocate swapspace when memory is low.
If you'd like to learn more about working and optimizing Vagrant then take a look at ADVANCED.md.
The file DEVELOP.md contains some tips for developers.
Thanks to Alex Holmes for the great work at (https://github.com/alexholmes/vagrant-hadoop-spark-hive)