wiki:GENIExperimenter/Tutorials/jacks/HadoopInASlice/ObtainResources

Hadoop in a Slice

Image Map

1. Establish the Environment

1.1 Ensure SSH keys are setup

Verify that you have at least one public key associated with your account. To do that, after you login to the portal check under your Profile, under the SSH keys tab. If you do not have SSH keys associated yet, please follow the instructions on that tab of the Portal.

1.2 Configure Omni

If you have not installed and configured omni yet, please follow these instructions.

2. Obtain Resources

2.1 Create a slice Slice

Create a slice using omni and the slice name of your choice. From now on that slice name will be referred to as SLICENAME.

$ omni createslice SLICENAME 

2.2 Create your RSpec

We are going to use the graphical interface of the portal (Jacks) to create the RSpec file for this tutorial but we are going to use Omni for reserving the resources.

2.2.1 Load a simple topology in Jacks

  1. In the Portal, open the Slice page for the slice you just created. Notice that you created the slice with omni and it is available via the Portal.
  2. Press the Add Resources button to launch Jacks for this slice.
  3. From the Choose RSpec menu , select from the drop down menu the rspec named: EG Hadoop Nodes
  4. After you click Select, a couple of nodes will appear.

The topology you loaded has two node-types: master and worker. Let's see what each node-type is comprised of:

  1. Inspect the properties of the hadoop master. Note that not only the usual attributes (e.g. node name, node type) are set, but there are also custom OS Image, install and execute attributes. These attributes is what customizes a generic GENI node a specialized one like a hadoop master.

2.2.2 Create the topology(RSpec) for the experiment

Now we will create our Hadoop cluster. For the cluster we need:

  • 1 hadoop master
  • 2 or more workers
  • all the nodes need to be on the same Layer 2 Domain and IP Subnet (172.16.1.0/24)
So let's go ahead and draw our topology:
  1. Select the worker node.
  2. Press the "Duplicate Nodes only" button.
  3. Repeat the above two steps so that you have at least three workers but no more than six (please be respectful to other and limit your cluster size, especially if you are doing this in the context of a tutorial).
  4. Delete the original worker node (select it and on the attributes on the left click the Delete button).
    DO NOT press the delete/backspace key since this is equivalent to pressing the Back button on the browser.
    Delete worker node
  5. Ensure that all your workers are named in the pattern `worker-i` with i=0,1.2,... (e.g. `worker-0, worker-1, worker-2`)
    WARNING This step is important, if the workers are not named based on the above convention your cluster WILL NOT be configured correctly
  6. On the master node make sure that on the Execute script the input number matches the number of workers, e.g. if you have 3 workers the entry should be
    /home/hadoop/hadoop_config_dynamic.sh 3
  7. Connect all your nodes in one LAN
    1. Drag a line between two of your nodes
    2. From each unconnected node draw a line from that node to the middle of the link where there is a small square Draw multipoint VLAN
  8. Assign IPs to your nodes:
    Click on the small square in the middle of your lan and set the IPs and netmask for each interface according to this list:
    • the netmask for all interfaces should be 255.255.255.0
    • master IP: 172.16.1.1 Master IP
    • worker-i IP: 172.16.1.<10+i>, e.g. worker-0: 172.16.1.10, worker-1: 172.16.1.11, worker-5: 172.16.1.15, etc
      WARNING: Make sure the IPs are assigned according to the above pattern, your cluster will not be configured correctly otherwise. Worker IP
  9. Inspect your topology to ensure that all the configurations are correct and Download your rspec.
    NOTE: The rspec will be saved in your default dowload folder (usually ~/Downloads) under a name similar to SLICENAME_request_rspec.xml

2.3. Reserve your resources

You can use any tool to reserve this topology, today we are going to use Omni. To do that you will need to:

  1. Submit the request by running this omni command in a command line window:
    omni createsliver SLICENAME RSPEC_FILE -a AM_NICKNAME
    where AM_NICKNAME is the nickname for your assigned aggregate manager, SLICENAME is the name of the slice you created earlier and RSPEC_FILE is the filename of the RSpec you just saved.
    If all goes well Omni should print a line close to the else like
    Result Summary: Got Reserved resources RSpec from ....

2.5 Wait for resources to be ready

You can tell whether your nodes are ready by using a script built on omni called readyToLogin.

  1. Please use the command:
     
    readyToLogin slicename --useSliceAggregates 
    
     
    where  slicename is your slice name
      
  2. If it reports that the sliver is not yet ready (for example, it might say that the status is "changing"), then wait a minute or two and try again. Once everything is complete, readyToLogin will give output that should look something like this:
     
    ... 
    master's geni_status is: ready 
    User vthomas logs in to master using:
    	ssh -i /Users/vthomas/.ssh/geni_key_portal vthomas@165.124.159.76
    	ssh -i /Users/vthomas/.ssh/geni_cert_portal_key vthomas@165.124.159.76 
    ... 
    

Introduction

Next: Execute the Hadoop Experiment

Last modified 3 years ago Last modified on 05/26/17 16:23:40

Attachments (24)