Adam Williamson <[EMAIL PROTECTED]> writes:

> On Wed, 2003-06-25 at 04:44, Leon Brooks wrote:
> > On Wed, 25 Jun 2003 11:02, Adam Williamson wrote:
> > > It's the urpmi --parallel functionality that's being referred to, I
> > > think. That's been around for a while. Read the manpage...
> > 
> > Ooh! Shiny! (-:
> > 
> > However, it doesn't explain very much about setting up the URPMI 
> > "slaves". Do you just give the "master" DSA-keyed root logins on the 
> > "slaves", or what?
> 
> I've no idea, I've never used it :). Fpons?

Well, there is a lack of "simple&good" documentation here, so I will add an
extra documentation now ;)

The format is as follow :
  <alias_name>:<extension_name>:<specific options>
or
  <alias_name>:<extension_name>(<specific_media_list>):<specific options>

Currently there exists only two extension, one with ssh and one with ka-run.

For ssh, specific_options is just a ':' separated list of nodes, each node
should already have been installed with urpmi and ssh-server correctly
configured (this means server keys of root should have been exported, or you
will have to type root password very often).
Example :
  leia:ssh(cooker):leia
which allow me to update pixel's machine :-) (just for test of course).
Another one :
  test2:ssh:dhcp125:dhcp119
which declare two nodes, dhcp125 and dhcp119.

For ka-run, typically used specific_options is as follows :
  -c ssh -m 192.168.100.200 -m 192.168.100.201 -m ...
where -c ssh allows us to use ssh with ka-run and -m <ip> indicates all nodes,
you may want to use machine names instead of ip of course.

ka-run extension are by far the most powerfull because the underlying protocol
(ka-run) is much faster for large cluster (as this tools is from CLIC or
Mandrake Cluster project) as it is optimized to copy files through multiples
nodes very efficiently.

I hope it is fine now,
Francois.

> -- 
> adamw

Reply via email to