Grid Appliance on FutureGrid - Eucalyptus
From Grid-Appliance Wiki
Contents |
Pre-requisites
Creating Account And Using Educational Appliances on FutureGrid
Introduction
This is an introductory tutorial for using grid-appliance images tailored for use on FutureGrid resources with Eucalyptus.
Getting Started
- Once you have received the required credentials to access a cloud, follow the FutureGrid - getting started with Eucalyptus tutorial up to the point where you are comfortable deploying an instance with the euca-run-instances command.
- To deploy grid-appliance instances on Futuregrid, you can either use our pre-packaged image or create your own custom image. If you want to use an existing image, our grid-appliance VM images have already been uploaded to a shared repository on the India cloud. Using the below commands, you should be able to see all available grid-appliance images on Futuregrid.
euca-describe-images | grep grid-appliance
- This command should output a list of images similar to the following: (The output is truncated for display purpose)
IMAGE emi-A5521473 grid-appliance-test/lucid-ga-test2.manifest.xml panoat available ... IMAGE emi-936F1441 grid-appliance-test/lucid-ga-test.manifest.xml panoat available ... IMAGE emi-6C5616D7 gridappliance/grid-appliance-hadoop-2.04.11.manifest.xml renato ... IMAGE emi-C8011472 grid-appliance/grid-appliance-2.04.17.manifest.xml davidiw available ... IMAGE emi-DEF61593 grid-appliance/grid-appliance-static.manifest.xml davidiw available ... IMAGE emi-BB2E1432 gridappliance/grid-appliance-2.04.11.manifest.xml renato available ... IMAGE emi-8D5213F6 grid-appliance/lucid-server-amd64.manifest.xml davidiw available ... IMAGE emi-FB2D111E grid-appliance/ubuntu-10.04.manifest.xml davidiw available ...
- Please note the "emi" number (The number on the 2nd column) of the image you want to use. This number will be used to deploy grid-appliance instances as described in the next section.
Deploying a Grid appliance connected to a public Condor pool
- You can deploy a Condor-based Grid appliance with the following command while replacing "emi-XXXXXXXX" with the actual number from the list. Note: the argument given for -k is name of the previously created Eucalyptus user keypair you want to use to log into the machine.
euca-run-instances -k userkey -n 1 emi-BB2E1432
- Once the instance reaches "running" state, its hostname/IP address is displayed when you run:
euca-describe-instances
- To log into the Grid Appliance, ssh using the root user (replace public.ip.address with the address given by Eucalyptus). Note: the argument given for -i is the file name with the private key associated with the keypair used to run the instance (in this example, userkey.private).
ssh -i userkey.private root@public.ip.address
- The appliance does not connect automatically to any pool at start-up when it is instantiated using the cloud-client. To connect to the default public pool, you need to simply install the grid-appliance-public-pool package in your grid-appliance.
apt-get install grid-appliance-public-pool
- Wait a few moments and check that your Grid appliance is connected to the Condor pool:
condor_status
Deploying a Grid appliance connected to your own pool using GroupVPN
To connect to any other GroupVPN pool, you need to place a floppy disk containing the GroupVPN configuration files inside the appliance in the /opt/grid_appliance/etc directory and then restart the grid_appliance service. This floppy can be generated using the grid-appliance.org web front-end. The steps are as outlined below:
- Stop the grid_appliance service in the grid-appliance
/etc/init.d/grid_appliance.sh stopexit
- Download the GroupVPN floppy to your local system from the grid-appliance website. Transfer it to the Grid Appliance workspace using scp or sftp.
scp -i userkey.private ./floopy.img root@public.ip.address:/opt/grid_appliance/etc/
- SSH back into grid appliance and start the grid appliance service
ssh -i userkey.private root@public.ip.address/etc/init.d/grid_appliance.sh start
- NOTE: for a Condor pool, you need at least to configure at least one machine with the "server" floppy image to run the job scheduler; the remaining machines should be "workers". Repeat steps 2 and 3 for as many nodes that you have deployed in your pool.

