[172187] in North American Network Operators' Group
Re: ipmi access
daemon@ATHENA.MIT.EDU (Brian Rak)
Mon Jun 2 16:00:11 2014
X-Original-To: nanog@nanog.org
Date: Mon, 02 Jun 2014 15:54:36 -0400
From: Brian Rak <brak@gameservers.com>
To: shawn wilson <ag4ve.us@gmail.com>, Nikolay Shopik <shopik@inblock.ru>
In-Reply-To: <CAH_OBieEv6f5sw-B4--zsiwCO5Hh0yu5BYqSn+yfGAo_AY=H8A@mail.gmail.com>
Cc: North American Network Operators Group <nanog@nanog.org>
Errors-To: nanog-bounces@nanog.org
On 6/2/2014 3:47 PM, shawn wilson wrote:
> On Mon, Jun 2, 2014 at 3:19 PM, Nikolay Shopik <shopik@inblock.ru> wrote:
>
>> Java only used for mouting images. KVM is transfered via VNC protocol iirc.
> They're not re-inventing the wheel, but I think KVM is generally some
> VNC stream embedded in http(s) which VNC clients can't seem to
> understand (at least, at a glance, I haven't been able to connect to
> iLo, DRAC, Spider, or Tyan IPMI from outside the Java app).
No, at least on SuperMicro it's a hacked up VNC protocol. It's not
embedded in HTTP/HTTPS, it just uses HTTP/HTTPS to fetch the Java app.
I say hacked up because it's got a custom auth method, and a whole bunch
of undocumented extensions. I looked into implementing support in noVNC
for it, but reverse engineering a binary protocol is a bit beyond me.
It's also annoying because it claims to be a TightVNC server (and uses
TightVNC auth/tunneling)... I was so hopeful that would just work. It
looks like they took the TightVNC code, and just made a bunch of changes
with no regard for the specification.
