UNIX for Beginners Questions & Answers

Hello,

sorry if this is an easy thing, but I tried to make it run the whole morning, but it doesn't work so far...

I am logged in on a visualization node of a high performance cluster. On this node I can run several gui-based software for the postprocessing of my cfd-data.

As the visualization node is always very busy, I would like to do the compute nodes the work.
My idea was to submit a bash script where I tell the compute nodes to start the specific software and to forward the display to my local display.

I read out the adress of my display with

Code:

echo $DISPLAY

and tried to start the software on the compute node with the following:

Code:

export DISPLAY="Adress of my display"
"Command for software"

As this did not work, I also tried this:

Code:

ssh -X myusername@server; set $DISPLAY="Adress of my display", "Command for software"

In both cases no X-window opened on my screen.

As I am a noob in unix I would be happy about any advice.

Thanks very much!

No problem. In fact there are no silly questions, just silly answers. Welcome on board.

OK, let us start with some background inforamtion to make you understand better what you are doing.

X-Windows is a completely networked system from the start. Unlike Windows where you have a local system onto which networking capabilities were added later, X-Windows is built for networked computers from the start.

It consists of three different parts: X-clients are regular programs which use the services of an X-server (see below) to put output onto a "display". X-server is a piece of software which manages some physical hardware (monitors, graphics cards, ...) to create such "displays" from it and offer its services over the network. There is a special protocol, X-protocol, for X-clients to communicate with X-servers over the network.

It is at first usually a bit confusing (but, see above, makes perfect sense) that X-clients run usually on the server side of things while X-servers run on the client. Picture an X-Server as a graphics card driver with a network interface.

Now, there may be serveral X-servers on a network and there may be several systems with running X-clients which use the various X-servers to display their output on the DISPLAY.

Which X-server to use for that is controlled via the environment variable DISPLAY. The DISPLAY-variable consists fo the IP-address of the system the X-server runs on (or a DNS-name leading to such an IP-address) and a number, because every X-server may manage several displays and every display might consist of several screens:

Code:

DISPLAY=1.2.3.4:0.0
        ======= = =
           |    | |
           |    | V
           |    | screen number within the display
           |    V
           |  Display number to use
           V
        IP-address the X-server runs on

So much for background, let us get to your real problem:

Code:

ssh -X

This is correct, because -X enables the X-protocol to be used over the communication channel opened by ssh.

set $DISPLAY="Adress of my display", "Command for software"

This is close, but still wrong. First, to set the DISPLAY variable and make it known throughout all processes started in there it must be "exported", second, there is a syntax error in your statement. Here is the corrected version:

Code:

export DISPLAY=<"addr":0.0> ; <your xclient to start> &

If this produces an error like "Unable to open display" you either have set the DISPLAY variable incorrectly or your X-client (to be precise: the system your X-client runs on) is not allowed to use the X-server you specified. To allow this there is a tool xhost in UNIX/Linux environments, but if you have an X-server running on a Windows system (like Opentext Exceed, Xming and various other products) there is probably some configuration item. You need to find out this, but you can probably search for "access control" or similar in the documentation to find that.

I hope this helps.

bakunin