Several issues with 396.24 under CentOS 7

Ok, so it is the “exec gnome-session” and not the x11vncservers in teh startx-script which gave me the impression that it hangs (ehile in fact it is just not existing from the gnome session, right)?

I did it npw with startx, and when I connect with VNC I get this (and no VNC screen comes up):

“Connection reset by peer” - does this mean I have some firewall issue on the vnc client side?

26/08/2018 16:01:58 Got connection from client 10.194.144.190
26/08/2018 16:01:58 other clients:
26/08/2018 16:01:58 Normal socket connection
26/08/2018 16:01:58 Disabled X server key autorepeat.
26/08/2018 16:01:58 to force back on run: ‘xset r on’ (3 times)
26/08/2018 16:01:58 incr accepted_client=1 for 10.194.144.190:52994 sock=13
26/08/2018 16:01:58 Client Protocol Version 3.8
26/08/2018 16:01:58 Protocol version sent 3.8, using 3.8
26/08/2018 16:01:58 rfbProcessClientSecurityType: executing handler for type 1
26/08/2018 16:01:58 rfbProcessClientSecurityType: returning securityResult for client rfb version >= 3.8
26/08/2018 16:01:59 created xdamage object: 0x1000002
26/08/2018 16:02:00 client useCopyRect: 10.194.144.190 0
26/08/2018 16:02:00 client_set_net: 10.194.144.190 0.0029
26/08/2018 16:02:00 Pixel format for client 10.194.144.190:
26/08/2018 16:02:00 32 bpp, depth 24, little endian
26/08/2018 16:02:00 true colour: max r 255 g 255 b 255, shift r 16 g 8 b 0
26/08/2018 16:02:00 no translation needed
26/08/2018 16:02:00 Using compression level 1 for client 10.194.144.190
26/08/2018 16:02:00 Enabling X-style cursor updates for client 10.194.144.190
26/08/2018 16:02:00 Enabling full-color cursor updates for client 10.194.144.190
26/08/2018 16:02:00 Enabling cursor position updates for client 10.194.144.190
26/08/2018 16:02:00 Using subsampling level 0 for client 10.194.144.190
26/08/2018 16:02:00 Enabling LastRect protocol extension for client 10.194.144.190
26/08/2018 16:02:00 Enabling NewFBSize protocol extension for client 10.194.144.190
26/08/2018 16:02:00 Using tight encoding for client 10.194.144.190
26/08/2018 16:02:04 rfbProcessClientNormalMessage: read: Connection reset by peer
26/08/2018 16:02:04 client_count: 0
26/08/2018 16:02:04 Restored X server key autorepeat to: 1
26/08/2018 16:02:04 viewer exited.
26/08/2018 16:02:04 deleted 1 tile_row polling images.

Looks rather like a bug/incompatibility in the vnc viewer:

26/08/2018 16:02:04 rfbProcessClientNormalMessage: read: Connection reset by peer

Can you check with a different viewer like ultravnc/tigervnc?

Yes I can connect with UltraVNC viewer but I just get a black screen (with a responsive mouse cursor).
The output on the server side is

26/08/2018 21:15:00 Got connection from client 10.194.144.190
26/08/2018 21:15:00   other clients:
26/08/2018 21:15:00 Normal socket connection
26/08/2018 21:15:00 Disabled X server key autorepeat.
26/08/2018 21:15:00   to force back on run: 'xset r on' (3 times)
26/08/2018 21:15:00 incr accepted_client=1 for 10.194.144.190:53635  sock=13
26/08/2018 21:15:00 Client Protocol Version 3.8
26/08/2018 21:15:00 Protocol version sent 3.8, using 3.8
26/08/2018 21:15:01 rfbProcessClientSecurityType: executing handler for type 1
26/08/2018 21:15:01 rfbProcessClientSecurityType: returning securityResult for client rfb version >= 3.8
26/08/2018 21:15:01 Pixel format for client 10.194.144.190:
26/08/2018 21:15:01   32 bpp, depth 24, little endian
26/08/2018 21:15:01   true colour: max r 255 g 255 b 255, shift r 16 g 8 b 0
26/08/2018 21:15:01 no translation needed
26/08/2018 21:15:01 rfbProcessClientNormalMessage: ignoring unsupported encoding type Enc(0x0000000A)
26/08/2018 21:15:01 rfbProcessClientNormalMessage: ignoring unsupported encoding type zlibhex
26/08/2018 21:15:01 Using compression level 6 for client 10.194.144.190
26/08/2018 21:15:01 Enabling full-color cursor updates for client 10.194.144.190
26/08/2018 21:15:01 Enabling cursor position updates for client 10.194.144.190
26/08/2018 21:15:01 Using image quality level 6 for client 10.194.144.190
26/08/2018 21:15:01 Using JPEG subsampling 0, Q79 for client 10.194.144.190
26/08/2018 21:15:01 rfbProcessClientNormalMessage: ignoring unsupported encoding type xorEnable
26/08/2018 21:15:01 Enabling LastRect protocol extension for client 10.194.144.190
26/08/2018 21:15:01 Enabling NewFBSize protocol extension for client 10.194.144.190
26/08/2018 21:15:01 rfbProcessClientNormalMessage: ignoring unsupported encoding type cacheEnable
26/08/2018 21:15:01 rfbProcessClientNormalMessage: ignoring unsupported encoding type Enc(0xFFFF8000)
26/08/2018 21:15:01 rfbProcessClientNormalMessage: ignoring unsupported encoding type Enc(0xFFFF8001)
26/08/2018 21:15:01 rfbProcessClientNormalMessage: ignoring unsupported encoding type Enc(0xFFFF8004)
26/08/2018 21:15:01 rfbProcessClientNormalMessage: ignoring unsupported encoding type Enc(0xFFFF8002)
26/08/2018 21:15:01 rfbProcessClientNormalMessage: ignoring unsupported encoding type Enc(0xFFFF8003)
26/08/2018 21:15:01 rfbProcessClientNormalMessage: ignoring unsupported encoding type Enc(0xC0A1E5CE)
26/08/2018 21:15:01 Using ZRLE encoding for client 10.194.144.190
26/08/2018 21:15:01 client useCopyRect: 10.194.144.190 -1
26/08/2018 21:15:01 client_set_net: 10.194.144.190  0.0028
26/08/2018 21:15:01 created   xdamage object: 0x1000002
26/08/2018 21:15:01 Pixel format for client 10.194.144.190:
26/08/2018 21:15:01   32 bpp, depth 24, little endian
26/08/2018 21:15:01   true colour: max r 255 g 255 b 255, shift r 16 g 8 b 0
26/08/2018 21:15:01 no translation needed
26/08/2018 21:15:01 rfbProcessClientNormalMessage: ignoring unsupported encoding type Enc(0x0000000A)
26/08/2018 21:15:01 rfbProcessClientNormalMessage: ignoring unsupported encoding type zlibhex
26/08/2018 21:15:01 Using compression level 6 for client 10.194.144.190
26/08/2018 21:15:01 Enabling full-color cursor updates for client 10.194.144.190
26/08/2018 21:15:01 Enabling cursor position updates for client 10.194.144.190
26/08/2018 21:15:01 Using image quality level 6 for client 10.194.144.190
26/08/2018 21:15:01 Using JPEG subsampling 0, Q79 for client 10.194.144.190
26/08/2018 21:15:01 rfbProcessClientNormalMessage: ignoring unsupported encoding type xorEnable
26/08/2018 21:15:01 Enabling LastRect protocol extension for client 10.194.144.190
26/08/2018 21:15:01 Enabling NewFBSize protocol extension for client 10.194.144.190
26/08/2018 21:15:01 rfbProcessClientNormalMessage: ignoring unsupported encoding type Enc(0xFFFF8000)
26/08/2018 21:15:01 rfbProcessClientNormalMessage: ignoring unsupported encoding type Enc(0xFFFF8001)
26/08/2018 21:15:01 rfbProcessClientNormalMessage: ignoring unsupported encoding type Enc(0xFFFF8004)
26/08/2018 21:15:01 rfbProcessClientNormalMessage: ignoring unsupported encoding type Enc(0xFFFF8002)
26/08/2018 21:15:01 rfbProcessClientNormalMessage: ignoring unsupported encoding type Enc(0xFFFF8003)
26/08/2018 21:15:01 rfbProcessClientNormalMessage: ignoring unsupported encoding type Enc(0xC0A1E5CE)
26/08/2018 21:15:01 Switching from ZRLE to ZRLE Encoding for client 10.194.144.190
26/08/2018 21:15:10 created selwin: 0x1000003
26/08/2018 21:15:10 called initialize_xfixes()
26/08/2018 21:15:21 cursor_noshape_updates_clients: 0
26/08/2018 21:15:22 cursor_noshape_updates_clients: 0

Might be a bug in mutter/gnome-shell. Please install openbox and replace in .xinitrc

exec gnome-session

with

exec openbox

The right mouse button should give you a menu then.

Edit: alternatively, you can also set a fake monitor in xorg.conf. Add to the ‘Device’ section

Option "ConnectedMonitor" "DFP-0"

When I use openbox, it is the same.

If I additionally use the option “Conncetd Monitor” in the Device or Monitor section, I get an error:

“ConnectedMonitor” is not a valid keyword in this section.

The error message
“ConnectedMonitor” is not a valid keyword in this section.
Points to forgetting the keyword
Option
Please check that the line you added is really
Option “ConnectedMonitor” “DFP-0”

Using openbox, you’ll get a black window with a mousepointer but right-clicking will show a menu.

Generix - THANK YOU SO MUCH!!! IT FINALLY WORKS AFTER ALMOST 2 YEARS!!!

The final mistake was that I missed the word “Option” in the config file.

How can I repay you??? I’m really indebted to you.

I just started MatLAB, and it is so much faster.

There are some minor issues still, but I guess that is some x11vnc / Linux configuration.

One thing is that I can’t resize windows (or only to a limited size). And I need to figure out how to overcome the locked console restriction (currently I do sudo touch /var/run/console , but I want to have it worked out for non-sudo users as well).

Finally, I’m not quite clear if several users need to run their own startx / x11vnc sessions, and how they are directed to “their” session when they login (does the client basically scan running sessions for users, and then the password is required to login?). To me it seems more logical if you have only one vnc-server running which handles several clients at the same time?

Again, thank you so much for your patience and perfect explanations. I never thought that I could make that run.

Don’t thank me too early…
This is only the most simple of all setups, takes about 5 minutes if you know what you’re doing.
While simple, this has also clear limitations. Since this is using a real Xserver running on a real gpu, this is limited to one (!) concurrent user since an Xserver needs exclusive access to the gpu.
So if you want more concurrent users to use hardware acceleration, this requires a different approach. Taking cost/time as measure, maybe Nomachine Workstation + VirtualGL would fit. Gives you accelerated access for up to 4 concurrent users. Taking into account that the video memory (4GB) of the Quadro has to be shared between users, I don’t think much more would be desireable.
A free implementation could be achieved using VirtualGL+TigerVNC+xrdp but this requires extensive setup.

To get a higher resolution, in your xorg.conf change
HorizSync 28.0 - 33.0
to
HorizSync 28.0 - 68.0
to have all user access the console, edit /etc/pam.d/xserver
commenting out
#auth required pam_console.so

Oh, I see. Thanks nonetheless, at least I know that the card and driver are fully working!

So If I understand correctly, it is the hardware acceleration (GLX) which - with the current setup - supports only one user at a time?

I’m afraid I actually don’t understand what an Xserver really does. I always thought that with a “regular” (or simple) graphic card (no GLX) you can of course run a VNC-server (as a part of the X-server) which supports several clients at a time - or is this not correct?

I looked at Nomachine workstation, I would not mind paying a little money if it works. However, the 4-user config (most simple setup) explicitly mentions

“Connect to hosts via RDP, VNC and XDM - N/A”

So this is why I additionally would need VirtualGL?

Yes, or better said, one Xserver at a time.

Partially. Simply put, an Xserver is a server that manages the resources of a graphics card (real or virtual) exclusively and lets clients (applications) connect to display something. Due to the architecture of X, you can actually run multiple virtual Xservers inside a real Xserver (Xephyr) making it possible for many users having their own xserver. This doesn’t work with GLX though.
A possibility to kind of split an Xserver is ZaphodHeads, making each monitor output of the graphics card its own screen. Fiddly to match the input devices to screens, though. You might find some ‘Linux MultiSeat’ configs on the web. Sidenote- :4 is just short for localhost:4.0 (host:xserver.screen)
VirtualGL is a different approach, it’s sort of a GLX broker, using an Xserver exclusively and rendering 3D content for multiple, even remote virtual Desktops.

No, NM uses its own client software, NX client for connections. Just the server version provide additional access though VNC, RDP…you should use the trial first to see if it meets your requirements.
VirtualGL is needed to have HW Accel on every virtual desktop.