Opening permissions on X displays under RHEL4/CentOS4

[wilville]

I have a network comprised of RHEL4 servers of X86_64 architecture and CentOS clients of i386 architecture. In X windows terms, the notion of client and server would be reversed, since the CentOS based machines are supplying the display hardware and the RHEL based machines are using them.

Now under RHEL3 I was able to place files in /etc named X0.hosts, X1.hosts and so on to allow machines listed in each file to use the corresponding display without having to issue an xhost + command for each. Under RHEL4/CentOS4, this seems to no longer work.

Can it be made to work, and if so how? Or is there a way to get the same result in a different manner?

 

[QatQat]

If I understood correctly you want some sort of terminal behavior.

Why not use LTSP?

QatQat

If I could have sex each time I reboot my server, I would definitely prefer Windoz over Linux!

 

[pansophic]

Have you run gdmsetup? From what I read you need to configure your xhosts in the Security tab in gdmsetup.

Optionally you can manually edit the /etc/X11/gdm/gdm.conf file and add:

DisallowTCP=false

Although it would be more secure, and works out of the box, if you use ssh -X instead.


pansophic

 

[wilville]

I actually need a little more than just terminal behavior. And, I agree fully with the sentiment about SSH. In fact either SSH -X and ssh -Y both work just fine, when they are applicable. BUT ( always a 'but' somewhere ), I am in the situation of running interactive jobs through a loac balancing tool called LSF. It has a similar, but slightly different, tunneling mechanism to ssh when run as interactive with shell services, which is what we are doing. When I had RHEL3 everywhere, I had opened up the display on each desktop to all the LSF cluster machines, since you never know where it will dispatch you by applying the X0.hosts, X1.hosts, etc files in /etc and running gdm to turn off DisallowTCP. That worked fine at that time. So when I cut over to CentOS4 ( read RHEL4 ), I did the same thing. And I did just check to see that the Disallow TCP setting was false on the desktop machines. Pleasingly. yes it is. But it still fails me. I worked so well under RHEL3 that I had my hopes up for RHEL4 as well. Thus my post.

However, now it comes to light that even if the user pays attention to where he got dispatched and runs xhost + for that server, it only works intermittently. Setting a DISPLAY variable and directing the display back to the opened up host appears to work solidly, but no one thinks to pay attention to where they get dispatched. So if I could open up to the whole cluster automatically, like I had it before, it would take a lot of frustration and handholding out of the process. I suppose I could run a script at short intervals to do a large xhost +, but that seems like a waste of good CPU cycles. The Xn.hosts solution was so clean and effective, that I hope it can be recoved.