Re: [gomd-devel] GOMD autodiscovery bug

[Gian Paolo Ghilardi]

Hi all (again).


> Hi Roel, me again,
>
> On Freitag 22 August 2003 11:53, address@hidden wrote:
> > Hi again, :)
> >
> > when executing "start ad 300", Gomd tries to reach IP adresses over 255
at
> > the end. It searched for 192.168.0.300, wich couldn't exist in IPv4.
>
> hehehehe, a good one ;)))
Wops.... :)))
Immediately fixed... :)
(see updated bugs_history file committed to CVS => bug #6)

>
> >
> > also, I think autodiscovery could be much easier and better using a
> > technoligy invented by Apple: "Rendezvous". this broadcasts services and
> > can search for services of a certain type. Howl is a package for Linux
> > implementing it. Get it at http://www.swampwolf.com/products/
>
> mhmmm, i do not like more deps ;)
I _hate_ deps... (hehehe...) ;)

>
> Anyways the question is why not simply use the local nodelist
> for the auto discovery ?
>
> The local gomd knows all system-id's from the active openMosix nodes
> in the cluster so there is (to my mind) no need to probe all ip addresses
> in the same subnet (also a system can have more then one nic, what happens
> then ?)
> To my mind we should derive the list with the ip-addresses to probe for a
> running gomd from the node-id's from the local nodelist.
> This will prevent probing a lot of ip-addresses which actually cannot run
> a gomd because they are non-openMosix nodes, not in the same cluster
> or offline.
> (for deriveing the ip-address from the node-id you can simply look into
> the omdiscd source or i can help)
>
As gomd can be run also on non-oM machines (in this case the daemon returns
system-only infos), the equation oM_node=running_gomd is not valid... ;)

Thanks to all for your great help!!!

Byez.

<rejected>