|
HPC Clustering quickstart for Red Hat Enterprise Linux 5 and Clones
(including but not limited to Centos and Oracle Unbreakable Linux)
[ note: looking for an all in one HPC Solution with modern kernel and drivers? Check out Caos Linux ]
To Begin, Go through a normal Enterprise Linux installation, making sure to do the following:
* Network IP address: Define at least one network device to be on an RFC1918 (address allocation for private networks). Use one of the following IP ranges:
10.0.0.0 - 10.255.255.255 (10/8 prefix)
172.16.0.0 - 172.31.255.255 (172.16/12 prefix)
192.168.0.0 - 192.168.255.255 (192.168/16 prefix)
* Hostname: Make sure you define a hostname that is not the default localhost/localdomain!
* Packages to select: It is recommended if you are installing an HPC cluster to unselect all of the package groups/tasks (the clustering group is for Loadballancing and High Availibility (HA) clusters, not HPC).
* Additional package groups: Depending on what packages you will require it maybe a good idea to include Development tools and libraries at this point (but not required). If you select "Customize Now" at the bottom of the page you can further customize the package selections.
Post installation configuration:
Disable firewall:
# /etc/init.d/iptables stop
# /sbin/chkconfig iptables off
Infiscale's EL YUM repository configuration: Thanks to our friends at AcceleratedServers.com for providing this new repo.
# wget -P /etc/yum.repos.d/ \
http://altruistic.infiscale.org/rhel/5/RPMS/infiscale.repo
Install and configure Perceus:
# yum install perceus
# perceus-init
Boot your nodes:
Power cycle each node in the order you wish for them to be numbered
within Perceus.
Install and configure Warewulf:
# yum install warewulf warewulf-tools
# /etc/init.d/warewulf restart
Install and configure Torque:
# yum install torque-server torque-scheduler torque-client
# /etc/init.d/pbs_server restart
# /etc/init.d/pbs_sched restart
# /etc/init.d/pbs_mom restart
Install of some misc Utilities:
# yum install pdsh pdsh-dshbak env-modules bash-completion
To add nodes to Torque, edit the server_priv/node file and list the nodes, or dump directly from Perceus assuming 4 CPU's per node:
# perceus -e node list | xargs -I '{}' -n 1 echo "{} np=4" \
> /var/spool/torque/server_priv/nodes
# /etc/init.d/pbs_server restart
|
