Monitoring Rocks Installations for Continuous Reboots

Submit New Article

May 6, 2010 8:00 PM PDT


Rocks cluster provisioning system, as most modern systems, does unattended installation. During the installation, Anaconda, the installation program used by Fedora, Red Hat Enterprise Linux, and some other distributions, makes a graphical installation.

However, the hardware drivers included in the distribution for video may be missing or conflict with the video card found in the system.  A common symptom of this failure is a continuous reboot cycle where compute nodes never complete their image installation.

The solution is to configure Rocks to install the nodes in "headless" mode. In this way, the graphical installation happens inside a VNC server and it can be monitored with a VNC client.

To configure Rocks to install a node in "headless" mode, run in the frontend, as root:

rocks set host installation <node name> action="install headless"

For Example:

rocks set host installation compute-0-0 action="install headless"

And then, reinstall (PXE-boot) the node.

To see the installation progress, in headless or normal mode, open an xterm in the frontend, and run, as root:

rocks-console <node name>

For Example:

rocks-console compute-0-0



Do you need more help?