On Tue, 9 Feb 2010, Jeremy Chadwick wrote:
On Tue, Feb 09, 2010 at 06:53:26AM -0600, Karl Denninger wrote:
Jeremy Chadwick wrote:
On Tue, Feb 09, 2010 at 05:21:32PM +1100, Andrew Snow wrote:
http://www.supermicro.com/products/motherboard/ATOM/ICH9/X7SPA.cfm?typ=H
Supermicro just released a new Mini-ITX fanless Atom server board
with 6xSATA ports (based on Intel ICH9) and a PCIe 16x slot. It
takes up to 4GB of RAM, and there's even a version with KVM-over-LAN
for headless operation and remote management.
Neat hardware. But with regards to the KVM-over-LAN stuff: it's IPMI,
and Supermicro has a very, *very* long history of having shoddy IPMI
support. I've been told the latter by too many different individuals in
the industry (some co-workers, some work at Yahoo, some at Rackable,
etc.) for me to rely on it. If you *have* to go this route, make sure
you get the IPMI module which has its own dedicated LAN port on the
module and ***does not*** piggyback on top of an existing LAN port on
the mainboard.
What's wrong with the Supermicro IPMI implementations? I have several -
all have a SEPARATE LAN port on the main board for the IPMI KVM
(separate and distinct from the board's primary LAN ports), and I've not
had any trouble with any of them.
http://unix.derkeiler.com/Mailing-Lists/FreeBSD/current/2008-01/msg01206.html
http://forums.freebsd.org/showthread.php?t=7750
http://www.beowulf.org/archive/2007-November/019925.html
http://bivald.com/lessons-learned/2009/06/supermicro_ipmi_problems_web_i.html
http://lists.freebsd.org/pipermail/freebsd-stable/2008-August/044248.html
http://lists.freebsd.org/pipermail/freebsd-stable/2008-August/044237.html
(Last thread piece does mention that the user was able to get keyboard
working by disabling umass(4) of all things)
I have a box down at Softlayer (one of the few major server rental outfits
that officially supports FreeBSD), and one of the reasons I went with them
is that they advertised "IP-KVM support". Turns out they run Supermicro
boxes with the IPMI card. It mostly works, but it is very quirky and you
have to use a very wonky Java client app to get the remote console. You
have to build a kernel that omits certain USB devices to make the keyboard
work over the KVM connection (and their stock FBSD install has it
disabled).
I can usually get in, but sometimes I have to open a ticket with them and
a tech does some kind of reset on the card. I don't know if they a
hitting a button on the card/chassis or if they have some way to do this
remotely. After they do that, I'll see something like this in dmesg:
umass0: <Peppercon AG Multidevice, class 0/0, rev 2.00/0.01, addr 2> on
uhub4
ums0: <Peppercon AG Multidevice, class 0/0, rev 2.00/0.01, addr 2> on
uhub4
ums0: 3 buttons and Z dir.
ukbd0: <Peppercon AG Multidevice, class 0/0, rev 2.00/0.01, addr 2> on
uhub4
kbd2 at ukbd0
The umass device is to support the "virtual media" feature that simply
does not work. It's supposed to allow you to point the ipmi card at an
iso or disk image on an SMB server and boot your server off of it. I had
no luck with this.
All the IPMI power on/off, reset, and hw monitoring functions do work well
though.
It gets worse when you use one of the IPMI modules that piggybacks on an
existing Ethernet port -- the NIC driver for the OS, from the ground up,
has to be fully aware of ASF and any quirks/oddities involved. For
example, on bge(4) and bce(4), you'll find this (bge mentioned below):
hw.bge.allow_asf
Allow the ASF feature for cooperating with IPMI. Can cause sys-
tem lockup problems on a small number of systems. Disabled by
default.
So unless the administrator intentionally sets the loader tunable prior
to booting the OS installation, they'll find all kinds of MAC problems
as a result of the IPMI piggybacking. "Why isn't this enabled by
default?" I believe because there were reports of failures/problems on
people's systems who *did not* have IPMI cards. Lose-lose situation.
I don't think they have this setup, or if they do, they are using it on
the internal LAN, so I don't notice any weirdness.
If you really want me to dig up people at Yahoo who have dealt with IPMI
on thousands of Supermicro servers and the insanity involved (due to
bugs, quirks, or implementation differences between the IPMI firmwares
and which revision/model of module used), I can do so. Most of the
complaints I've heard of stem from serial-over-IPMI. I don't think
it'd be a very positive/"supportive" thread, however. :-)
One similar product that does seem to work well is iLO, available on
HP/Compaq hardware.
I've heard great things about that. It seems like a much better design -
it's essentially a small server that is independent from the main host.
Has it's own LAN and serial ports as well.
Charles
--
| Jeremy Chadwick j...@parodius.com |
| Parodius Networking http://www.parodius.com/ |
| UNIX Systems Administrator Mountain View, CA, USA |
| Making life hard for others since 1977. PGP: 4BD6C0CB |
_______________________________________________
freebsd-stable@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-stable
To unsubscribe, send any mail to "freebsd-stable-unsubscr...@freebsd.org"
_______________________________________________
freebsd-stable@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-stable
To unsubscribe, send any mail to "freebsd-stable-unsubscr...@freebsd.org"
