[GE users] I need help badly.

Chris Dagdigian dag at sonsorol.org
Thu Mar 1 22:18:28 GMT 2007


Trey -

Lots of ways to replicate directories on Linux,  'rsync' is a great  
tool for doing this and it can run over an encrypted SSH tunnel which  
makes it useful to use across the internet or to internal machines  
that only have SSH exposed through a firewall.

This may work for you instead of 'scp'

On your qmaster host and running as root:

# rsync -avz -e ssh /sge root at centos2.hyper.com:/sge

That will do a recursive directory copy via SSH to the 2nd machine

That said though, much like the gentle encouragement you got from  
this list to use SGE 6.0 you'll find that things will be a lot easier  
to manage, administer and debug when you are sharing your SGE  
directory via NFS to your cluster nodes. In addition to making  
installation easier this also means that all your configuration data,  
spool logs and other core files are in a single easily-checkable place.

In addition to the sharing of SGE most people also NFS share the home  
directories of the users as well. From a user's perspective, things  
are much much easier when the home folder is shared and consistent  
cluster-wide. If you are going to go through the trouble of sharing "/ 
sge" or something then you'd also probably want to share out "/home/"  
as well.

Setting up NFS is not a big deal but you don't want to run any type  
of NFS server exposed to the internet. You referenced a hostname of  
"centos2.hyper.com" so I want to make sure you are doing this  
internally or over a private network first.   Since you are using  
Centos, just surf on over to the online manual located at http:// 
www.centos.org/docs/4/html/rhel-sag-en-4/ and check out the chapter  
on NFS.

-Chris





On Mar 1, 2007, at 5:00 PM, Trey wrote:

> OK  I have lost of my notes from when I was trying it on 5.3 (as I  
> switched to 6.0) and cant find the exact command to copy a whole  
> directory.  I think it was scp -r /sge/default  
> root at centos2.hyper.com /sge.  But that doesnt seem to work  
> correctly.  I would like to know how to set up the NFS on the linux  
> boxes for ease and so that I know how to do it (As you probably  
> already know, I am a windows guy trying to learn as I go and this  
> is a project that got thrown in my lap).
>
> Chris Dagdigian wrote:
>> Most people share $SGE_ROOT via NFS to all the nodes that are part  
>> of the Grid Engine cluster.
>> If you don't do this, you'll need to replicate $SGE_ROOT to the  
>> other node so that (as you have found) default/common is present.
>> And nope, you don't want to install the qmaster there.
>> -Chris
>> On Mar 1, 2007, at 4:42 PM, Trey wrote:
>>> when you install an execution host on another linux box, do I  
>>> have install the qmaster on all of them?  Default/commons is not  
>>> on any of them and the ./install_execd requires them.

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
For additional commands, e-mail: users-help at gridengine.sunsource.net




More information about the gridengine-users mailing list