summaryrefslogtreecommitdiff
path: root/parallel/glunix/MESSAGE
blob: 080394f8508d1d8032177e1c666302c2c74678b0 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
===========================================================================
$NetBSD: MESSAGE,v 1.2 2002/09/24 12:30:30 wiz Exp $

QUICK START GUIDE:

Change your directory to the now installation base, eg:
${NOW_ROOT}/bin.  Run ./glupart install as root.

You must initialize the partition management system to
understand the machines in your cluster.  The tool to use is
"glupart".  You will need to do a "glupart machine <name> new" for
each new machine.  Multiple machines can be added at a time using a
comma-separated list of names.  You can also do "glupart -help" for
more detailed information.

To run GLUnix, you need a single machine which is able to do a root
rsh (or ssh or kerberized rsh) without a password to all of the
machines you want to run GLUnix on.  Call this trusted machine the
"master" machine.  The scripts that get GLUnix running initially (the
run_glunix_* scripts) need to run an rsh command on each node in order
to set up GLUnix.  Log on to the master machine and run "run_glunix
-start <nodes>" where "<nodes>" is the list of nodes you want to run
GLUnix on:  generally a comma-separated list of machines.  (You can do
"run_glunix -help" to see all the options.)  If the names of some of
the machines are of the form of some base string followed by
consecutive numbers (our machines are u0, u1, u2, u3, u4), then these
machines can be abbreviated to the run_glunix command as "u0..u4".
Here is a sample run_glunix command:

	run_glunix -start thunk,sparc,u0..u110,pc0..pc9,dawn0..dawn30,cayuse

Once the run_glunix command has successfully completed, you can run
"glustat" to find out how many nodes came up.

===========================================================================