"beoserv" daemon not reacting to slave rarp requests ...

Daniel Ridge newt at scyld.com
Tue Nov 21 15:11:39 PST 2000


Richard,

On Tue, 21 Nov 2000, Richard Walsh wrote:
> "beoserv" (listening from eth0 on port 2223) and bpmaster 
> seem to come up without trouble. The config file is correct
> (eth0 is the internal address 10.0.0.17, eth1 is a second
> internal network for file transfers only, and eth2 is external)
> Range (10.0.0.1 10.0.0.16) and netmask are specified correctly.

> The slave nodes boot from floppy and begin r-arping.
> This is clear from the lights on the interfaces and
> a tcpdump on the head node. At this point, the config 
> file in /etc/beowulf/config should be updated ... but this 
> never happens ... the slave nodes just r-rarp forever. 

These nodes are actually going into '/var/beowulf/unknown_addresses'.

These are moved into '/etc/beowulf/config' via the 'beosetup' tool
or manually. (see comments in this file for details...)

Beyond that, the only suggestion I can make is that you might want
to have the frontend before your IP range instead of after. This
may make it slightly less traumatic to expand the size of your cluster
later.

> I am assuming that bpmaster and beoserv take care of all 
> that is required to assign the IP address out of the specified 
> range in the config file, but perhaps I am wrong ... does 
> some other service have to be running?

Run 'beosetup'. We don't presume -- entirely automatically -- that
you want a given box in your cluster. The beosetup tool can be 
configured to automatically add new nodes.

Good luck,
	Dan Ridge
	Scyld Computing Corporation





More information about the Beowulf mailing list