HP Insight Cluster Management Utility - Manual (Page 121)

The transfer uses TCP/IP sockets. The clone image is saved to the local disk. The node then asks
the image server if any successors are waiting for upload. If any successors are waiting, the node
then starts to transfer the image to a group member, while the image server uploads a third one.
This process is called the tree propagation algorithm.
After a node has received a completed image, it attempts to upload to another node within the
entity. This mechanism speeds up the propagation process and takes advantage of the available
network bandwidth. Each time a node receives a clone image, the node uncompresses the image
on the local disk. This is designed to speed up the cloning process. The process is also performed
on all secondary servers.
When the HP Insight CMU list of nodes to be cloned completes, each node within the group has
an identical system disk image on the local system disk. To reboot on the local disk, some node
specific files must be adjusted on a node-by-node basis, such as the TCP/IP node address and
hostname. HP Insight CMU adjustment is limited to the TCP/IP name and address of the primary
network interface. The TCP/IP hostname and IP address set for compute nodes is defined by the
DHCP configuration file on the image server, for example /etc/dhcpd.conf.
During cloning, secondary servers boot from the primary server.
Then, each compute node boots from its secondary server.
7.6 Cloning mechanisms 121