Re: diskless cluster
> It's a little bit complicated, but when it's all working, well worth it!
Complete agreement! And you can have a few tens os nodes booting off a
single 9 GB disk on the server. With logs and all, full read-write mounts.
A user entering from the net can't tell a node from a full-fledge system.
We use the same scheme for terminals scattered throughout the floor.
> To reboot a client, just switch it off and on.
Or use a serial console and reboot it, see it going down and comming up
again, including the Etherboot prompt, inside a nice X11 window from the
confort of you office! Cool, and useful if you hit some hard-to-figure
hardware or system software problem.
> To add a new one, just use update-cluster-add and
> update-cluster-regenerate all.
We have our own tools for that. Including, you can use Debian and update
or upgrade a node with the node off! Or all nodes, for that matter. All
action goes on in the server with the use of the chroot command. Very,
very cool. Also quite fast and very convenient.
Cheers,
----------------------------------------------------------------
Jorge L. deLyra, Associate Professor of Physics
The University of Sao Paulo, IFUSP-DFMA
For more information: finger delyra@latt.if.usp.br
----------------------------------------------------------------
Reply to: