Differences

This shows you the differences between two versions of the page.

--- archive_cluster [2010/01/12 23:34] – /* Configuration */ _marcoe
+++ archive_cluster [2026/09/27 04:46] (current) – removed - external edit (Unknown date) 127.0.0.1
@@ Line 1: / Line 1: @@
-====== Cabling ======
-Each box ([[cluster:sun|Sun]],[[cluster:mercury|Mercury]],[[cluster:venus|Venus]],[[cluster:earth|Earth]]) is plugged into the switch and has it's own IP on the cluster VLAN (.35,.36,.37,.38 respectively) this is connected the the first port on each box (for some reason eth2 within Debian although labeled eth0).
-The net-management ports for Mercury, Venus and Earth are plugged into the other three network ports of Sun. Use the following commands from sun to connect:
-  - Mercury: telnet 192.168.2.2
-  - Venus: telnet 192.168.2.4
-  - Earth: telnet 192.168.2.6
-====== Configuration ======
-TODO:
-- Setup PXE booting so that the other 3 nodes get up and running. Shared root from head node Sun.
-- /home should be distributed GlusterFS. Issue is when one or more nodes are down, /home will be offline. Find solution to problem.
-- User authentication via tardis' LDAP server. Require users to be in 'cluster' group.
-- Password-less ssh logins within cluster. Kerberos(preferred) or script to generate ssh public keys for every new user.
-- MPI. OpenMPI (preferred if matured enough) or MPICH2.
-- SSI. Kerrighed if sparc is supported. (conflicts with SGE)
-- Sun Grid Engine. Might be overkill for our small cluster. (conflicts with SSI)
-Last Modified: marcoe 23:25, 12 January 2009 (GMT)
-If possible, shutdown unused nodes via LOM.