Introduction:

Running a Rocks 6 Cluster on ProfitBricks.com is easy and robust.  ProfitBricks uses infiniband backend connections that provide an ideal cluster setting for a small scale cluster or a development cluster.  Here are a few steps to get you started.

Setup your new DC on the Dev cluster so that you can keep costs low and use the latest hardware and software from PB. https://devops.profitbricks.com/signup/

You will want to log into your Data Center Designer and create a new Data Center.

 rocksdcd

  1. Upload Rocks 6.1 DVD via FTP first so that it will be available when you get ready to install.  ftp://ftp.rocksclusters.org/pub/rocks/rocks-6.1.1/linux/area51+base+bio+fingerprint+ganglia+hpc+htcondor+java+kernel+kvm+os+perl+python+sge+web-server+zfs-linux-6.1.1.x86_64.disk1.iso
  2. Request a dedicated IP from the IP Manager link at the top of your DCD.
  3. Add a server with sufficient RAM (Minimum 8G) and CPU (Minimum 4)
  4. Add storage block with a minimum 50G initial storage device as a head node, you can add an additional data store later if necessary.
  5. Add 2 network adapters from the right side of the DCD control panel.
  6. Add an internet block to your DCD.
  7. Connect eth1 to the internet and choose your requested dedicated IP in the dropdown.
  8. Uncheck dhcp and gateway on eth0.
  9. Add CD-ROM to your new server and select the Rocks DVD ISO you uploaded and select boot from the right panel under storage.
  10. Provision your changes.
  11. Open a Remote Console to your server and type “build” at the first prompt once the DVD boots.
  12. Enter your IP address and a netmask of 255.255.255.0 with a gateway of your ip with .1 at the end then add dns 8.8.8.8 and 8.8.4.4 (Google)
  13. On initial login, check that the MTU of both interfaces is set to 64000 using a root terminal: (#  rocks set network mtu private 64000 )( # rocks set network mtu public 64000 ) ( # rocks list network ) ( # rocks sync config )
  14. Edit /etc/sysconfig/iptables to allow remote access to your new cluster.
  15. In the DCD, remove the image from the CD-ROM on the server.
  16. Reboot your server.
  17. Once the system is back up, open an SSH terminal session to your new clusters public IP address.
  18. Check that the MTU has been set to 64000 ( cat /etc/sysconfig/network-scripts/ifcfg-eth0 ) ( cat /etc/sysconfig/network-scripts/ifcfg-eth1 )
  19. As root ( # insert-ethers ) then select compute.
  20. Provision a new server with CPU, RAM, and at least 21G of storage.
  21. Connect the network of your new compute node eth0 to eth0 on the head node. Be sure to uncheck the DHCP box on each new node under the Network tab.
  22. Rinse and repeat as many times necessary to get the total number of CPU’s and RAM you will need.
  23. Once the new compute node comes online, it will pxe boot from the head node and you will see the mac address pop up on your root terminal session.  It will then install automatically and show up in your Ganglia monitoring page.  http://yourpublicip/ganglia/