Creating Head Node and Worker Node Images
From Debian Clusters
Contents |
Head Node
The head node is the machine your users will interact with. Unlike the worker nodes, users can SSH and run commands on this node. This means it may need some extra functionality that the worker nodes do not, and vice versa.
Two tweaks to consider making on the head node include
Worker Nodes
Setting up one worker node and then cloning the other nodes with that one's image can save a lot of time. Rather than having to set up each one for NFS, DNS, and LDAP, one machine can be set up and all the configurations are copied over when the others are imaged. Before cloning the hard drive, make sure you've set up all the services that will be running on the worker nodes. (The exception to this is any scientific software that will be installed later on an NFS mount.
There are also a few optional steps.
Renaming Network Interfaces
The network interfaces (normally eth0, eth1, etcetera) can be renamed if you so desire. However, this will need to be done individually on each machine after cloning. In addition, the entries for the existing MAC addresses should be cleared out before cloning the hard drive.
This is done through Udev.
Hostnames through DNS
Rather than manually editing /etc/hostname on each worker, nodes can be set to automatically assign their names based on DNS. See Setting a Dynamic Hostname by DNS.
Cloning
After the one machine has been set up correctly, the other worker nodes can then be cloned using that machine's image. See Cloning Worker Nodes.
