Re: configuration?

[Hans Schou]

On Wed, 30 Nov 2011, Ole Tange wrote:

> On Tue, Nov 29, 2011 at 5:20 PM, Alan McKay <alan.mckay@gmail.com> wrote:
>
> > I am the Sys Admin in a scientific research lab, and one of the
> > scientists asked me to install parallel onto our compute cluster.
> >
> > That was the easy part thanks to yum :-)
>
> Good to learn.
>
> > But I don't see any documentation at all on how to configure it.  How
> > do I tell it what other machines are out there to use?

I think he has a bunch of servers - which is growing - and alle users 
who have an account on the 'master', can use a pre defined list of 
servers.

Example: The user login to node00.example.com and then he has access 
to use node01.example.com to node07.example.com . He needs to have his 
own account as this is where he store his own files. Now he need 
access to all the nodes which is available today. A simple way to do 
that is by having a centralized list on node00, say in 
/etc/parallel/sshloginfile and all users whould just type:

 parallel --slf /etc/parallel/sshloginfile --nonall echo \$HOST

In this case the Sys Admin has to communicate to all users that there 
is this special file and it is beeing maintained (you can complain 
about it).

If it instead becomes a part of parallel then all users would read 
about it in the parallel man-page and look in 
/etc/parallel/sshloginfile to see if list is there.

Suggestion for change to '--sshloginfile':
==============
The sshloginfile '.' is special, it read sshlogins from 
/etc/parallel/sshloginfile

The sshloginfile '..' is special, it read sshlogins from 
~/.parallel/sshloginfile
==============

Another solution could be that '..' implies first 
/etc/parallel/sshloginfile but will get overruled by 
~/.parallel/sshloginfile if it exist.


I can see two situations where /etc/parallel/sshloginfile might be 
changed.
 1. Adding a new node
 2. Taking one node out for maintenance
        (capasitors dry out sometimes)

/hans
--
Horsebakken 78, DK-2400 København NV