Configure EMC Isilon for HDFS

EMC Isilon is a scale out NAS solution that supports CIF, NFS and native HDFS. There is a ton of information out there about why Isilon is cool but for HDFS it really address some issues with the “commodity model”. By having a central repository for our HDFS data, we don’t need to have multiple distributed copies of data as a traditional HDFS model does. We don’t have to use HDFS tools to import data into the cluster, we can just use NFS of CIFS, or perhaps the data is already on our Isilon cluster. We can do things like replicate data, dedupe (hadoop dedupe), or allow easy access for other applications to the data without having to import/export. Bigger hadoop clusters spend much of their life just ingesting data, and Isilon can seriously trim that time down! Add in that it’s a free license and simple setup and there seesm to be no reason not to use it!!!

This blog will show you how to configure you EMC Isilon array for use by HDFS in hadoop environments.

Logon to your Isilon cluster
To add HDFS license click the help button in the top right corner and select “About This Cluster” HDFS is a Free license avalaible from Isilon
Click Activate License and add code. After you should see the HDFS module listed NOTE: HDFS is a free license. You can obtain your code from your Isilon sales team.
Next click on File System Management and File System Explorer. In the right panel highlight the root /ifs By default HDFS has access to the root directory. You can change this at the command prompt. To create a specific directory for HDFS Click the Add Directory button in the middle of the page. You can also share this directory as an NFS export or CIFS share for easy ingestion of data for Hadoop
Enter the name “Hadoop” Give a user rights to the directory Click Submit NOTE: You can create a specific user to access the HDFS directory or you can use root. The command to create a user form the CLI is isi auth users create --name="user1"
You will now see the Hadoop directory in the root
Create a NFS export of this directory using Unix Sharing under the Protocols tab. Click “Add Export”
Enter information for the share Enter the path to the directory recently created At the bottom of the page click Save
Verify successful creation of export
SSH into the Isilon cluster. Run the command: isi hdfs This verifies that HDFS is running and shows the root directory that HDFS will use
Change the HDFS root to /ifs/Hadoop by running this command isi hdfs –root-path=/ifs/Hadoop Next, run: isi hdfs Executing this command again verifies results
Another option is to change block size. Block sizes can be up to one gig. The type of data you will be analyzing with hadoop will determine the block size to use. This example shows changeing the block size to 1 GB isi hdfs --block-size=1GB Next run isi hdfs to verify   Most distributions use the user mapred for jobtraker to access HDFS. To create that user and add him to the wheel group follow this step. SSH into the isilon cluster The mapred user needs temp space on HDFS when map jobs are run. When I tested Apache, Horonworks and Pivotal the /tmp folder is automatically created the first time a job is run and the mapred user is given permissions. For Cloudera we need to setup the /tmp structure and give it permissions Below is a screen shot showing the setup done from the CLI of an Isilon cluster:   Note: /ifs/Hadoop is the HDFS root.

Comments

[higkoo]

Last picture can use JUST two command :
# mkdir -pv /ifs/Hadoop/tmp/hadoop-mapred/mapred/{staging,system}
# chown -R mapred:wheel /ifs/Hadoop/tmp

[Nesa]

How does Isilon addresses copying petabyte of data from storage to data nodes for processing? And how it adresses doing this several times a day? Or doing this for multiple jobs in the same time? This data has to go through network?