Baracus was going to be awesome.

No, not a team. Unless you think of a team of people working on a project to make your network install of that latest release easier. Or, unless maybe you're thinking of a team trying to manage the way your systems boot.

Baracus: "Boot and build served your way." Well, ok, yeah, it's the tag-line, and it's descriptive of what the core of the project is about.

Baracus is about building systems, from vendor released ISOs. Baracus also provides for deploying some snarfed up disk image, or a programmatically produced "appliance" disk image from say SuSE Studio, or one you like made by someone else shared in the corresponding SuSE Gallery. And Baracus allows users to craft lots of customizations into those builds and do some interesting things following image deployment as well.

Baracus is about creation and use of network install sources for openSUSE, Fedora, Ubuntu (both desktop and server), and dare they mention support for more commercial distributions like Novell's SUSE Linux Enterprise, Red Hat Enterprise Linux, and dun dund daaaaaa... Windows 7 and Windows Server 2008. And for all of these you can craft your install with native install formats, autoyast, kickstart, preseed,and unattended script. This tool will help manage those scripts, and more.

Baracus is about boot management using the glorious and powerful gPXE project's undionly images (non-NIC rom). The beauty of the gPXE project (that was once called etherboot, and is now disappointingly forked into gPXE and iPXE) is that it layers so much goodness over PXE boot. And with that goodness they avoid most all the TFTP choking on it's own spew with timeouts and other unreliableness (and replace it with DHCP net0 unreliableness that I'm sure the gPXE folks hope to work out). What you do get with gPXE is http, and iSCSI and AoE and just generally more networking choices for your PXE booting pleasure. Hey, the project is liked by the KVM folks so much they use it for their BIOS/NIC 'network' bootloader.

Baracus with gPXE allows for some really interesting directions to be taken with every node PXE boot. And that, with some custom workloads, is where the boot management comes in, with tasks like hardware inventory collection (currently using HardWare LiSter), disk wipe from the powerful DBAN and more.

Baracus is coming. Baracus is going to make some serious system administrators out of otherwise ordinary users. Hummmm, maybe that's the new tag-line.

Disclaimer: This is a personal opinion, vaguely informative, post about a defunct open source project called Baracus, and is not meant for any commercial endorsement or influence on decision making.

Binary Drivers Pain and Suffering

So, you run Linux, have a fairly modern Nvidia card, and a flat panel display that supports more than 1280x1024, but only supports 1.x versions of EDID. Well, if you'd like to use native resolution you'll need the binary driver from Nvidia. Even versions of xorg as recent as 7.1 have a known bug because there is no work aournd for the incorrect "Panel Size" being reported. (See your /var/log/Xorg.0.log file.)

Fetch the binary driver from Nvidia. For instance I'm running SuSE Linux Enterprise Server 10 on an x86_64 arch... So, I grab Linux AMD64/EM64T at the time of this writing it's Latest Version: 1.0-9746

You'll want to install that (running a non-XEN kernel - as root - and do 'telinit 3' first) by building it - follow the guide. (There are other ways to point YAST install sources at nvidia drivers - but they're out of date - and you'll want to do it this way for what's comming.) After this you should be able to 'telinit 5' and get back into X at your display's native resolution. Sweet!

Mostly eye candy, but some workflow benifits are available with XGL (Desktop Effects) turned on.... Go to the Control Center and do that. It's purdy.

But wait, you say... "Pansy, my special problem... I must run XEN!" So, after seeing that you have a beautiful big display and lots of pretty things to look at you reboot into your kernel-xen dom0 and the desktop manager trys to launch but X dies over and over and you're left at a text prompt about to cry.

There's a patch to build this nvidia.ko for your XEN dom0... I think post #10 is where this OpenSuSE write-up came from. Original patch credits to 'honny' subsequently modified and latest version posted by 'JaXXon'.

This patch: patch-nv-1.0-9625-xenrt.txt applies fine to the Nvidia driver 1.0-9746 - and almost builds fine (follow the OpenSuSE write-up) with three exceptions.

• Make sure you have installed kernel-source, kernel-, kernel-xen, and kernel-syms
• Remove the 'xen-sanity-check' from the NV*/usr/src/nv/Makefile 'module: ' directive
• Use SYSSRC=/lib/modules/2.6.16.27-0.6-xen/source in the make invocation.\

And that, should be all there is to that.

Enjoy

GNOME - Gconf, NFS $HOME, orbitrc tricks

So, for the longest time I wondered why using Gnome as a desktop for a user (with an NFS automounted $HOME directory) on two machines would cause all sorts of silly warnings and error messages from Gnome applications - related to GConf.

Two reasons.

The first is that the Gnome IPC mechanisim is only good on the localhost unless you tweak an dotfile in your $HOME directory. The file is '.orbitrc' and it should contain the following two lines

ORBIIOPIPv4=1
ORBIIOPUSock=1

The second is that gconf puts lock files in your home dir and somethimes those dangle after a crash or, well, I really don't care. All I know is that I've had a look at this document (GConf configuration system)and found out how to shut gconfd down and remove the lock files.

gconftool-2 --shutdown
rm ~/.gconfd/lock/ior
rm ~/.gconf/%gconf-xml-backend.lock/ior

All the information here is in that document.... I just wanted to make it quick for me to remember the next time my remote gnome-terminal comes up with crazy fonts and colors.

ssh -X <machinename> 'bash -l -c "gnome-terminal \ --window-with-profile=<profilename>"'

Xorg -nolisten tcp

So what's the point of Xauth and DISPLAY settings and xhost + if it all boils down to a radio button in a "Login Screen" control panel?

I suppose it's great that these Linux distros (like fedora core 6 (test 3)) are now shipping with all security options set to paranoia level 'RED' but damn it make it easier to turn this garbage off for those of us who like living on the edge.

On windows, X clipboard selection, vnc, and cut-and-paste

I use a VNC client on MS Windows XP to connect to a VNC server on a Linux host -- and the general lack of ability to cut and paste to/from the Windows applications from/to the Linux X applications via VNC i s more than bothersome!!!

I have had this problem with both RedHat 9 and RHEL4.

Looking for background I found a fredesktop.org write up on the behavior of the text selection and its being copied to the clipboard buffer as described by ICCCM for X applications.

This set of keywords lead me to find autocutsel -- an application meant to manage changes for the clipboards of Windows and X via the VNC client. You should run this application as a service on X startup - invoking 'autocutsel -fork'.

Still, I know for certain that the X clipboard was not managed by the RealVNC server (version 4.1.2 was the last tried - they may have fixed it since) as needed for autocutsel. So, finally, I changed the server on the RHEL4 box from the default (version number looks like its from RealVNC) to the VNC server provided by TightVNC.

All is right again in the world. Cut-and-paste Copy-and-paste from/to X to/from Windows all works as expected. Yeah!

At some point maybe i'll try something that others have buzzed about in the past: Synergy.

Python and LibGlade

I've found that Python and LibGlade are also wonderfully paired these days. This Python code

import gtk
import gtk.glade

class appGui:
  def __init__(self):
      callbacks = {
          "on_button1_clicked"        : self.on_button1_clicked,
          "on_window1_delete_event"   : gtk.main_quit
          }
      self.do_bindings(callbacks)
      return

  def do_bindings(self, callbacks):
      self.xml = gtk.glade.XML("simple.glade")
      self.xml.signal_autoconnect( callbacks )
      return

  def on_button1_clicked(self, widget):
      entry = self.xml.get_widget("entry1")
      print entry.get_text()
      return

appGui()
gtk.main()

uses the same simple.glade file that the Overholt demo created. Try it!

Fedora Core 5 and Dual Head or rather MergedFB

With FC5 and X11R7 things have changedfor Dual Head.

'man radeon' was *extreamly* helpful. Real documentation.

For the ATI Radeon driver on an IBM ThinkPad T30, I left out "Xinerama" and "Clone" from the "ServerLayout" section and added the following to the "Device" section:

Option "MonitorLayout" "LCD, VGA"
Option "MergedFB" "True"
Option "CRT2Position" "LeftOf"
Option "MetaModes" "1024x768-1280x1024"
Option "MergedNonRectangular" "True"

Also in the "Display" SubSection of the "Screen" Section added

Virtual 2304 1024

So, all is well after 'pkill -HUP X11' except that I had to move panels and icons back to the laptop LCD as I intend that to be the primary. I didn't try "MergedFB" "Auto".