If someone is doing this, especially wrt to 1U servers, perhaps someone could provide a pointer to a reboot solution that would allow a single reboot device to control 42+ 1Us in a single rack. From my occasional interest in this area, I've seen lots of reboot solutions. Most of them take as many as 4-10 U to reboot 40 machines, which sort of obviates the need to reboot 40 machines in a single rack in the first place. Unless you use 10ft racks, I guess.
IMHO the right way to do this is to build the power cycling capability into the individual 1U boxen. Again, we should be talking to the embedded systems folks to give them a standard set of requirements that everyone can support. This may be worth having a workshop discussion at a NANOG meeting to define requirements. I envisage a standard control module with two external serial ports. One serial port is for a daisy-chained serial bus that can allow one control module to talk to other boxen. The other serial port is where the modem connection goes and could probably be designed to take either an RJ11 phone cable or an RJ45 serial terminal cable. The 1U box's power supply has a built-in relay that is switched by the control module. All external switches and indicators go through the control module. --Michael Dillon
On Mon Mar 22, 2004 at 12:11:04PM +0000, Michael.Dillon@radianz.com wrote:
One serial port is for a daisy-chained serial bus that can allow one control module to talk to other boxen.
Personally, I have a major dislike for daisy-chained control buses like this. All it needs is one box at the start of the bus to go bad and you've lost control all the other boxes on the bus. Simon -- Simon Lockhart | Tel: +44 (0)1628 407720 (x(01)37720) | Si fractum Technology Manager | Fax: +44 (0)1628 407701 (x(01)37701) | non sit, noli BBC Internet Ops | Email: Simon.Lockhart@bbc.co.uk | id reficere BBC Technology, Maiden House, Vanwall Road, Maidenhead. SL6 4UB. UK
On Mon, 22 Mar 2004 Michael.Dillon@radianz.com wrote: :: IMHO the right way to do this is to build the power cycling :: capability into the individual 1U boxen. Again, we You've probably seen it already, but nonetheless: http://www.rackable.com/products/compute.htm -jba __ [jba@analogue.net] :: analogue.networks.nyc :: http://analogue.net
:: IMHO the right way to do this is to build the power cycling :: capability into the individual 1U boxen. Again, we
You've probably seen it already, but nonetheless:
this is IPMI. tons of motherboard manufactures do this today (i just don't know which colo companies offer this). -- hanulec@hanulec.com cell: 858.518.2647 && 516.410.4478 https://secure.hanulec.com EFnet irc && aol im: hanulec
google "cmbus" On Mon, 22 Mar 2004 Michael.Dillon@radianz.com wrote: I envisage a standard control module with two external serial ports. One serial port is for a daisy-chained serial bus that can allow one control module to talk to other boxen. The other serial port is where the modem connection goes and could probably be designed to take either an RJ11 phone cable or an RJ45 serial terminal cable. The 1U box's power supply has a built-in relay that is switched by the control module. All external switches and indicators go through the control module. --Michael Dillon --matt@snark.net------------------------------------------<darwin>< Flowers on the razor wire/I know you're here/We are few/And far between/I was thinking about her skin/Love is a many splintered thing/Don't be afraid now/Just walk on in. #include <disclaim.h>
participants (5)
-
jeffrey.arnold
-
just me
-
Michael Hanulec
-
Michael.Dillon@radianz.com
-
Simon Lockhart