Difference between revisions of "IRIDIA cluster installing cluster software"

From IridiaWiki
Jump to navigationJump to search
 
 
(One intermediate revision by the same user not shown)
Line 14: Line 14:
 
</pre>
 
</pre>
   
(All modifications done both on the master and on the clients -- TODO: HALVA WHAT DO YOU MEAN BY THIS?)
+
(All modifications done both on the master and on the clients -- TODO: HALVA WHAT DO YOU MEAN BY THIS, WHICH MODIFICATIONS DO "ALL" REFER TO IN "ALL MODIFICATIONS" --- I know it should be easy, but I am still slightly confused?)
   
 
Added to /etc/services:
 
Added to /etc/services:
Line 20: Line 20:
 
sge_commd 536/tcp # Sun Grid Engine
 
sge_commd 536/tcp # Sun Grid Engine
 
</pre>
 
</pre>
  +
  +
  +
Export the Sun Grid Engine directory. Modify
  +
<pre>
  +
/etc/diskless-image/config.sh
  +
</pre>
  +
in order to add the correct entry in /etc/fstab on each client node. Search for the block that starts with "cat > /etc/fstab << EOF" and add two (TODO: I only see one!) lines below:
  +
<pre>
  +
192.168.100.101:/usr/local/sge /usr/local/sge/ nfs rw,defaults, 0 1
  +
</pre>
  +
  +
Add the line to:
  +
<pre>
  +
/usr/lib/diskless-image/template/etc/fstab
  +
</pre>
  +
as well.
  +
  +
Rerun dpkg-reconfigure diskless-image-simple, then create the directory:
  +
<pre>
  +
/usr/local/sge
  +
</pre>
  +
in the diskless image.
  +
  +
Then run
  +
<pre>
  +
update-host-directories.
  +
</pre>
  +
  +
Now on majorana:
  +
<pre>
  +
cd /etc; ln -s . rc.d; cd /usr/local/sge/;
  +
./install_qmaster
  +
</pre>
  +
  +
On the diskless image: copy rcsge to /etc/init.d/ and then run:
  +
<pre>
  +
update-rc.d rcsge defaults 95
  +
update-host-directories
  +
</pre>
  +
  +
On polyphemus: add an entry to /etc/fstab to mount r01:/usr/local/sge and
  +
<pre>
  +
cp /usr/local/sge/default/common/rcsge /etc/init.d; update-rc.d rcsge defaults 95
  +
</pre>
  +
  +
'''Note:''' SGE is on majorana, but the schedules is still on polyphemus!

Latest revision as of 14:00, 17 February 2005

This page contains information on installing clustering software such as the Sun Grid Engine and parallel programming middleware such as PVM and MPI.

Sun Grid Engine

The Sun Grid Engine can be installed from the backup or from scratch. Below are the instructions on installing it from a backup:

Copy the backed up tar.bz2 on majorana and untar it into /usr/local/sge. Add the following line to /etc/profile:

. /usr/local/sge/default/common/settings.sh

and in /etc/csh.cshrc:

source /usr/local/sge/default/common/settings.csh

(All modifications done both on the master and on the clients -- TODO: HALVA WHAT DO YOU MEAN BY THIS, WHICH MODIFICATIONS DO "ALL" REFER TO IN "ALL MODIFICATIONS" --- I know it should be easy, but I am still slightly confused?)

Added to /etc/services:

sge_commd	536/tcp	# Sun Grid Engine


Export the Sun Grid Engine directory. Modify

/etc/diskless-image/config.sh 

in order to add the correct entry in /etc/fstab on each client node. Search for the block that starts with "cat > /etc/fstab << EOF" and add two (TODO: I only see one!) lines below:

192.168.100.101:/usr/local/sge   /usr/local/sge/	nfs rw,defaults,	0	1

Add the line to:

/usr/lib/diskless-image/template/etc/fstab

as well.

Rerun dpkg-reconfigure diskless-image-simple, then create the directory:

/usr/local/sge

in the diskless image.

Then run

update-host-directories.

Now on majorana:

 
cd /etc; ln -s . rc.d; cd /usr/local/sge/;
./install_qmaster

On the diskless image: copy rcsge to /etc/init.d/ and then run:

update-rc.d rcsge defaults 95
update-host-directories

On polyphemus: add an entry to /etc/fstab to mount r01:/usr/local/sge and

cp /usr/local/sge/default/common/rcsge /etc/init.d; update-rc.d rcsge defaults 95

Note: SGE is on majorana, but the schedules is still on polyphemus!