Rolling a rack on carpet

Robert G. Brown rgb at
Fri Oct 25 13:53:10 PDT 2002

On 25 Oct 2002, Dean Johnson wrote:

> On Fri, 2002-10-25 at 10:27, Mark Andrew Smith wrote:
> > > like drive lights. Since you have the cord right there, you have no need
> > > for the power switch. It was just a matter of putting led's in slot
> > 
> > ...that is true for AT motherboards, but some nodes in my cluster use an ATX
> > motherboard and the on/off switch is on the front case.  I would imagine
> > most newer pieces of kit are ATX these days.....
> > 
> Yanking the power cord works with nearly all past and present machines.
> If you are patient and willing to wait for some machines internal UPS
> (our AT&T 3B20 has biga** truck batteries in the back) to run down, the
> percentage goes up. Of course it isn't really the best way, but it
> works. I have no problems with it because any computational nodes that
> get irrepairable software type problems from such an approach just get
> reinstalled with "Dean's Magic Disk" (read "aint kickstart great"). ;-)

And with ext3 filesystems, they almost never have irrepairable (or even
noticable) filesystem problems, although one can always lose files that
are in the process of being written.

ext3 rocks.  They don't even take that long to boot back up later, since
they no longer have to do that fsck thing if you kill power suddenly.

In our server room, though, we've got two-column racks that are bolted
to the floor around the power poles.  One has to be careful to maintain
enough clearance to be able to get at front (to be able to mount or
dismount the chassis from the rack) and rear (to cable them up) anyway.

So although you might not need to get at the front often, or (as Dean
notes) to reboot, being able to move it so you CAN get at the front or
designing your space so you don't have to move it at all is still a good
idea.  That way the whole discussion becomes moot.


Robert G. Brown	             
Duke University Dept. of Physics, Box 90305
Durham, N.C. 27708-0305
Phone: 1-919-660-2567  Fax: 919-660-2525     email:rgb at

More information about the Beowulf mailing list