Linux Hadoop Minimal Notes Version .42 Date: June 3, 2019 Author: Douglas Eadline (Email: deadline(you know what goes here)basement-supercomputing.com) Unless otherwise noted, all course content, notes, and examples are Copyright Basement Supercomputing 2019, All rights reserved. What Is This? ============= The Linux Hadoop Minimal is a virtual machine (VM) that can be used to try the examples presented in the two on-line course entitled: "Hands-on Introduction to Apache Hadoop and Spark Programming" "Practical Linux Command Line for Data Engineers and Analysts" It can also be used for the examples provided in the companion on-line video tutorial (14+ hours) "Hadoop and Spark Fundamentals: LiveLessons" The machine has many important Hadoop and Spark packages installed and at the same time tries to keep the resource usage as low as possible so the VM can used on most laptops. (See below for resource recommendations) To learn more about the course and my other analytics books and videos, go to: https://www.safaribooksonline.com/search/?query=eadline PLEASE NOTE: This version of Linux Hadoop Minimal (LHM) is still considered "beta." If you use it and find problems, please send any issues to deadline@eadline.org with "LHM" in the subject line. Student Usage: ============== If you have taken the "Hands-on" course mentioned above, you can download the NOTES.txt files, examples, and data archive directly to the VM using "wget" The archive is in both compressed tar (tgz) and Zip (zip) format. It is recommended that you either make a new user account or use the "hands-on" account for the archive (and run most of the examples from this account). For instance, to download and extract the archive for the "Hands-on" course from within the VM: wget https://www.clustermonkey.net/download/Hands-on_Hadoop_Spark/Hands_On_Hadoop_Spark-V1.5.tgz tar xvzf Hands_On_Hadoop_Spark-V1.5.tgz Similarly, for the "Linux Command Line" course (do this within the VM) wget https://www.clustermonkey.net/download/Hands-on_Hadoop_Spark/Linux-Command-Line-V1.0.tgz tar xvzf Linux-Command-Line-V1.0.tgz If you want to move files from your local machine to the VM, then you can use "scp" on your host. (scp natively available on Linux and Macintosh systems, it is part of the MobaXterm package on Windows) scp -P2222 SOURCE-FILE USERNAME@127.0.0.1:PATH USERNAME is a valid account on the VM. There is a user account called "hands-on" that can be used for most of the examples. Therefore the command to copy file (SOURCE-FILE) from your host system to the VM is: scp -P2222 SOURCE-FILE hands-on@127.0.0.1:/home/hands-on See the "Connect From Your Local Machine to the LHM Sandbox" below for more information on using ssh and scp. USAGE NOTES: ============ 1. The Linux Hadoop Minimal includes the following Apache software CentOS Linux 6.9 minimal Apache Hadoop 2.8.1 Apache Pig 0.17.0 Apache Hive 2.3.2 Apache Spark 1.6.3 Apache Derby 10.13.1.1 Apache Zeppelin 0.7.3 Apache Sqoop-1.4.7 Apache Flume-1.8.0 Spark 1.6.3 is installed because later versions need Python 2.7+ (not available in CentOS) 2. The Linux Hadoop Minimal has been tested with VirtualBox on Linux, MacOS 10.12, and Windows 10 Home addition. It has not been tested with VMware. 3. The Linux Hadoop Minimal Virtual Machine is designed to work on minimal hardware. It is recommended at a MINIMUM your system have 2 cores, 4 GB memory, and 70G of disk space. The VM is set to use 2.5G of memory. This will cause some applications to swap to disk, but it should allow the virtual machine to run on a 4GB laptop/desktop. (If you are thinking of using the Hortonworks sandbox then 4+ cores and 16+ GB of memory is recommended) 4. The above packages have not been fully tested although all of the examples from the course work. Installation Steps: ------------------- 1. Download and install VirtualBox for your environment. VirtualBox is freely available. Note: Some windows environments may need the Extension Pack. https://www.virtualbox.org 2. Follow the installation instructions for your Operating System environment. For Red Hat based systems this page, https://tecadmin.net/install-oracle-virtualbox-on-centos-redhat-and-fedora, is helpful. With Linux there is some dependencies on kernel versions and modules that need to be addressed. If you are using Windows, you will need an "ssh client." Either of these will work. They are both freely available at no cost. (MobaXterm is recommended) 1. Putty: http://www.putty.org (provides terminal for ssh session) 2. MobaXterm: http://mobaxterm.mobatek.net (provides terminal for ssh sessions and allows remote X Windows session) 3. Make sure hardware virtualization is enabled in your BIOS. 4. Download the https://www.clustermonkey.net/download/Hands-on_Hadoop_Spark/Linux-Hadoop-Minimal-0.41.ova image and load into VirtualBox. (NOTE newer version may be available.) 5. Start the VM. All the essential Hadoop service should be started automatically. Connect From Your Local Machine to the LHM Sandbox: --------------------------------------------------- It is possible to login and use the sandbox from the VirtualBox terminal, however, you will have much more flexibility with local terminals. Follow the instructions below for local terminal access. As a test, open a text terminal and connect to the sandbox as the root user with ssh. Macintosh and Linux machines have ssh and terminal installed, for windows see above (Putty or MobaXterm) or this document: https://www.clustermonkey.net/download/Hands-on_Hadoop_Spark/DOS-Linux-HDFS-cheatsheet.pdf The root password is: hadoop ssh root@127.0.0.1 -p 2222 You are should now be in the /root directory To confirm all the Hadoop daemons have started enter "jps" as root The results should list the 10 daemons as shown below. (process numbers will be different) # jps 1938 NetworkServerControl 2036 ZeppelinServer 1797 ResourceManager 1510 NameNode 1973 RunJar 1576 SecondaryNameNode 1882 JobHistoryServer 1675 DataNode 1962 RunJar 1841 NodeManager 2445 Jps Copying Files Into and Out of the Virtual Machine (VM) ------------------------------------------------------ To copy a file from your LOCAL MACHINE into the VM, use the "scp" command. For instance, to copy the file "SOURCE-FILE" from your local directory on your LOCAL MACHINE to the "hands-on" account. The password is "minimal" and the command places file in /home/hands-on directory in the VM. scp -P2222 SOURCE-FILE hands-on@127.0.0.1:/home/hands-on To be clear, the above command is run on your LOCAL MACHINE. On Macintosh and Linux systems run this from a terminal. On Windows run it from MobaXterm. To copy a file from the VM to your LOCAL MACHINE and place it in your current directory use the following. (don't forget the "."): scp -P2222 hands-on@127.0.0.1:/home/hands-on/SOURCE-FILE . To be clear, the above command is run on your LOCAL MACHINE. On Windows, the data will be placed in the MobaXterm "Persistent Home Directory." In the case of Windows 10 with user "Doug" this would be the following: C:\Users\Doug\Documents\MobaXterm\home Adding Users: ------------- As configured, the LHM comes with one general user account. The account is called "hands-on" and the password is "minimal" You can run everything under this account (but remember you need to be user "hdfs" to do any administrative work in HDFS. The hdfs account has no password. To become the hdfs user, log in as root and issue a "su - hdfs" command. Warning: Running as user "hdfs" gives you full "root" control of the HDFS file system. To add yourself as a user. a) As root do the following to create a user and add a password: useradd -G hadoop USERNAME passwd USERNAME b) These steps change to user hdfs and create the user directory in HDFS (as root) su - hdfs hdfs dfs -mkdir /user/USERNAME hdfs dfs -chown USERNAME:hadoop /user/USERNAME exit c) Logout and login to the new account Web Access: ----------- The various web interfaces shown in class are available using the following URLs. Enter the desired URL in you local browser and the VM should respond. HDFS web interface: http://127.0.0.1:50070 YARN Jobs web Interface: http://127.0.0.1:8088 Zeppelin Web Notebook: http://127.0.0.1:9995 The Zeppelin interface is not configured (i.e. it is run in anonymous mode without the need to log-in). The "Zeppelin Tutorial/Basic Features" notebook used in class works as does some of the SparkR notebooks. The "PySpark Example" that was demonstrated in class also works. Also, the "md" and "sh" interpreters have been tested and work. Loading Data into Zeppelin: --------------------------- If you want to load you own data into a Zeppelin notebook, place the data in the zeppelin account under /home/zeppelin. Login as root to place data in this account then change the ownership to zeppelin user for example: # cp DATA /home/zeppelin # chown zeppelin:hadoop /home/zeppelin/DATA This location is the default path for the Zeppelin interpreter (run "pwd" in the %sh interpreter). World Database for Sqoop Example: --------------------------------- MySQL has been installed in the VM. The World database used in the Sqoop example from the class has been preloaded into MySQL. Log Files: ---------- There is currently no logfile management and log directly may fill up and use the sandbox storage. There is a clean-logs.sh script in /root/Hadoop-Minimal-Install-Notes/Hadoop-Pig-Hive/scripts/ This script will remove most of the Hadoop/Spark and system logs (somewhat aggressive) Stopping and Starting the Hadoop Daemons: ----------------------------------------- The Hadoop Daemons are started in the /etc/rc.local file (the last script file that is run when the system boots) The actual scripts are in /usr/sbin and are very simple with no checking. If you are knowledgeable, you can check /var/log/boot.log for errors and issues. The scripts are run in the following order /usr/sbin/start-hdfs.sh /usr/sbin/start-yarn.sh /usr/sbin/start-derby.sh /usr/sbin/start-hive-metastore.sh /usr/sbin/start-hiveserver2.sh /usr/sbin/start-zeppelin.sh A corresponding "stop script" is run when the system is shutdown or rebooted. As mentioned, if all the the scripts are running, the "jps" command (run as root) should show the following (process numbers will be different) The RunJar entrees are for the hiveserver2 and hive-metastore processes # jps 1938 NetworkServerControl 2036 ZeppelinServer 1797 ResourceManager 1510 NameNode 1973 RunJar 1576 SecondaryNameNode 1882 JobHistoryServer 1675 DataNode 1962 RunJar 1841 NodeManager 2445 Jps For HDFS to be running correctly the following daemons need to be running: NameNode SecondaryNameNode DataNode If one or all are not running, run (as root) /usr/sbin/stop-hdfs.sh /usr/sbin/start-hdfs.sh For YARN to be running correctly the following daemons need to be running: ResourceManager JobHistoryServer NodeManager If one or all are not running, run (as root) /usr/sbin/stop-yarn.sh /usr/sbin/start-yarn.sh A local metadata database (called Derby) is needed for Hive, if the "NetworkServerControl" daemon is not running, then stop and restart the derby daemon: /usr/sbin/stop-derby.sh /usr/sbin/start-derby.sh So that Spark can use Hive tables there is hive-metastore and hiveserver2 service needed. To stop and restart the services (in the following order) /usr/sbin/stop-hiveserver2.sh /usr/sbin/stop-hive-metastore.sh /usr/sbin/start-hive-metastore.sh /usr/sbin/start-hiveserver2.sh Finally, if the Zeppelin web page cannot be reached, the Zeppelin daemon may no be running. Stop and restart the daemon: /usr/sbin/stop-zeppelin.sh /usr/sbin/start-zeppelin.sh If any or all of the daemons will not start after the above procedure then the is a bigger issue with the VM. Please contact Eadline and describe the situation. Stopping the VM --------------- To stop the VM, click on "machine" in the VirtualBox menu bar. Select "Close" and then select the "Save State" option. The next time the machine starts it will have all the changes you made. Alternatively, you can enter (as root user) from within the VM: # poweroff And the VM will gracefully shutdown the Hadoop/Spark services and preserve any changes you made. VM Installation Documentation ----------------------------- Please see /root/Hadoop-Minimal-Install-Notes directory for how the packages were installed. Issues/Bugs ----------- 1. Either create your own user account as described above or use the existing "hands-on" user account. The examples will not work if run as the root account. 2. If zip is not installed on your version of the VM, you can install it by entering the following, as root, and a "y" when asked. Zip will now be installed and available for use. # yum install zip Loaded plugins: fastestmirror Setting up Install Process [...TEXT ...] Total download size: 259 k Installed size: 804 k Is this ok [y/N]: y [...TEXT ...] Installed: zip.x86_64 0:3.0-1.el6_7.1 3. In previous versions there is a permission issue in HDFS that prevents Hive jobs from working. To fix it, perform the following steps: a) login to the VM as root (pw="hadoop") ssh root@127.0.0.1 -p 2222 b) then change to hdfs user su - hdfs c) fix the permission error: hdfs dfs -chmod o+w /user/hive/warehouse d) Check the result hdfs dfs -ls /user/hive e) The output of the previous command should look like: Found 1 items drwxrwxrwx - hive hadoop 0 2019-01-24 20:43 /user/hive/warehouse f) Exit out of the hdfs account exit g) exit out the root account exit You should now be back at the terminal on your laptop/desktop