Hi, There's a linux cluster somewhere (using Rocks), and I use remote desktop (nxmachine) to do the work on the frontend of the cluster. The desktop is Gnome. And, I intend to install new nVidia driver, but it requires me to stop the X Server. I check around the internet, they say we can use the /etc/init.d/gdm stop, but I can't run it since the gdm file does not exist. Another alternative is by rebooting the Linux (of course, we change the inittab before). But I don't know how to reboot remotely. Moreover, is it safe to reboot Linux remotely, because I can't access physically to the linux workstation. Oh, the operating system of the Linux is CentOS 4. What should I do ? Thanks.

Use SSH on a command line terminal

# ssh -l root <Linux box IP>
# init 3
# sh ./NVIDIA-XXXXX.bin
# init 5

Remember to make the proper changes to the xorg.conf file and if you cannot log in as root directly, use a different user and then su or sudo the commands.

If you're running a firewall on the remote machine, allow connections on port 22 at least while you install the driver, to reboot you can use ... drumroll ... "reboot" on the CLI, but if you have access to X, there should be a shutdown menu item that allows you to reboot.

Hi,
Thanks for the info, I tried the step similar to yours in :
http://wiki.centos.org/HardwareList/Nvidia_Graphics?highlight=(X+Server)

I used PuTTY to log in as root
-telinit 3 -> aftar telinit 5, the home folder of my regular user (/home/EonStrife/). But I thought it's ok, maybe I can get back after telinit 5
-sh NVIDIA*.run -> Installed smoothly
-telinit 5 -> Nothing happens ! the home folder /home/EonStrife/ does not return !

What happen ? How can I recover the home folder of /EonStrife/ ?

Thanks.

Maybe you can't access that folder because you're a different user. Try this command when you are logged in as root:

su EonStrife

Hi,
Not really because of that, here's what I got if I su EonStrife from root
:

[root@cluster ~]# su EonStrife
bash-3.00$ _

if I start a new connection using PuTTY using EonStrife account, it will complain that it can't found the /home/EonStrife/ folder and want to create new encryption keys (id_rsa.pub, id_rsa, authorized_keys) like it's the first time I login into the system using EonStrife

Thanks.

You must have done something else, there should be no correlation at all between installing the driver and disappearance of user folders, did you misuse "rm"?

If the X server doesn't come back on after init/telinit 5, you should read the NVIDIA readme and change what is necessary on /etc/X11/xorg.conf, you have to tell X to use the offcial nvidia driver as opposed to the open source one there.

OR maybe since there is no gdm on that machine you have to bring it up manually with startx.

Hi,
No...I didn't issue command 'rm'...

this is the bash history:

sh NVIDIA-Linux*.runlscd ..cd homecd EonStrife/Installers/sh NVIDIA*.run                -> try to install, but got error message/etc/init.d/gdm stop          -> try stopping using gdm, but couldn't find it/etc/init.d/gdm stopstopx                        -> if there's startx, then there should be stopx...lscd ~              -> I issued telinit 3 around here, but strangely the commandcd /              -> isn't listed in bash historycd ..ls                -> starting getting panic because home folder disappearscd homelslscd ..lscd homelscd condor        -> what's with the /home/condor/ folder ???lscd ~lssh NVIDIA*.run  -> never mind, install the driver firsttelinit 5       -> and return to previous statelscd /cd ..lscd home        -> but /EonStrife/home/ disappears !ls             -> panic, try various thingsexittelinit 3cd cd/lscd ..cd homecd EonStrife/Installerscd EonStrifelscd ..slscd homelslscd ~lscd ..lscd homelsexittelinit 5

Sorry, seems like whatever happened, happened before you started that sequence of commands, I really cannot help you any further, you could had created the user without a home directory and didn't noticed, deleted the directory before trying to install the driver, could be logging in a different machine, the machine could had been wiped and reinstalled, etc.

Only you can figure out what happened there.

thanks.
I create a new account and start anew.
Curious though, even though I've installed the newest nVidia Driver (with CUDA support), when I use glGetString(GL_RENDERER) and glGetString(GL_VERSION), I got:

Version 1.2 (1.5 Mesa 6.4.1)
Renderer Mesa GLX Indirect

but when I check the xorg.conf, it's already correctly set to use nVIDIa implementation...

Section "Module"    Load           "dbe"    Load           "extmod"    Load           "type1"    Load           "freetype"    Load           "glx"EndSection...Section "Device"    Identifier     "Videocard0"    Driver         "nvidia"    VendorName     "NVIDIA Corporation"    BoardName      "GeForce 8800 GTX"EndSection

What's wrong here ?

Thanks.

Did you restart the Xserver?

Try nvidia-xconfig to generate a proper xorg.conf file, though it looks fine. The driver installer should had removed the symbolic links to Mesa.