Connection to Xdisplay looses

[RAVI B. HIREMATH]

Hi gurus and pals,

Well here we are having a strange problem. We have following
configuration for our new application server:
HP-UX 11.00
3 CPU (750MHz clock speed)(PA 2.0 8700)
4GB RAM
Model: L3000(rp5400)

We use the x-display of the application UI of above said server
on a different machine, which is running 10.20 HP-UX. We use the following method:
1. telnet
2. export DISPLAY=:0
3. run application UI.

After sometime, if there is no transaction of any bytes through this term,
then the term hangs, doesn't accept any input and all the application UIs
are also hang(like frozen; if minimise and restore them, they
will be blank).

I surely beleive, there must be some soultion for this problem. If you need
some more info on this, please reply.

thanks in advance,
Ravi B Hiremath

[Bill Hassell]

First, make sure that both machines are up to date on patches. Xwindows has a number of components so you'll need to add all of them (and dependencies) for a complete solution. If this does not fix the problem, you probably have a network problem. What is between the two machines? Routers? Switches? Firewalls? Routers and firewalls may be denhying communication over the Xwindow socket although typically this is a all or nothing setting. If the network is overloaded, or if this is a wide-area network, there may be delays or packet losses that are causing the protocol to hang.

[RAVI B. HIREMATH]

Hi Bill,

Thanks for your reply. But there is no network problem
between application server(HP-UX 11.00) and xdisplay machine(HP-UX 10.20).
Ping works fine with the same application server from the xdisplay
machine(without any packet loss/delay.

Here it is WAN network and with other HP-UX xdisplay machine also
have the same problem with the application server. We tried a lot to
debug the server, but so far no luck. It is at
a customer place, and we're
in a tough condition now, b'coz of this.

I suspect there might be some patch for inetd or telnetd.... or some change
must be applied to a configuration file. If you
can suggest, I would appreciate.

thanks and regards,
Ravi B. Hiremath

[RAVI B. HIREMATH]

Hi all,

One more info, thought this might be some clue for the problem:
When tried 'rlogin' to the application server, to display the
application UI, this also meets the same fate. But interestingly this
can close the connection(unlike the telnet term, which just freezes),
with the following message on the term:
-----------------------------------------------
rlogin: reader: recv: Connection timed out
Connection closed.
-----------------------------------------------

Now one more update is, we have upgraded the HP-UX version on the
xdisplay machine, from 10.20 to 11.00. Even with this update, there is
no change in the behaviour.

regards,
Ravi B Hiremath

[Bill Hassell]

Since it sounds like the application server is connected tro a WAN, this is very likely the problem. ping is useless in evaluating network problems other than major connectivity issues. The fact that this problem is with rlogin, telnet and Xwindows confirms the problem.

WAN connections can be quite difficult to troubleshoot without special tools. Using ping is like calling the phone company about modem problems and they say something like: Can you hear me? Yes? Then the phone line is fine."

Without a network trace, you won't be able to see the low level datagram handshaking. HP-UX has the nettl program to provide the needed details. There may be some ndd parameters that you can set to help with the problem connection but first, you'll need the trace to provide the details. The nettl trace may indicate that the problem is with your WAN provider, firewall or routers,

[kish_1]

Can you change the MTU to 1450 and try on both server , I had simillar kind of problem , our n/w people did the same and they could solve the issue

[RAVI B. HIREMATH]

Hi Bill,

Good suggestion. We have nettl running on the server. But would you
tell me, which trace entity must be added in the /etc/nettlgen.conf file
and trace to be started to which entity.

thanks & regards,
Ravi B Hiremath

[RAVI B. HIREMATH]

Hi Babu,

I think I can change the MTU parameter by the following command:
# lanadmin -M 1440 0

Am I right? And this has to be done on the server as well as
xdisplay(HP-UX) machine, is it?.. And do I need
to reboot after this change
or it gets changed dynamically? Please elaborate as I've never used
it earlier.

thanks & regards,
Ravi B Hiremath

[kish_1]

you are corect, not required the reboot

[RAVI B. HIREMATH]

Hi Babu,

Currently on the both the application server and the xdisplay machine
it is set to 1500. I would like to know that, would that become better
if I change it to 1450(or 1440) as you suggested. I would like
to know, how it differs between 1500 and 1450(or 1440) and how it can
improve my situation.

rgds,
Ravi B Hiremath

[kish_1]

MTU specifies the maximum transmission unit size of an interface. Each interface used by TCP/IP may have a different MTU value specified. The MTU is usually determined through negotiation with the lower driver, using that lower driver value. However, that value may be overridden.TCP to attempt to discover the Maximum Transmission Unit (MTU or largest packet size) over the path to a remote host. By discovering the Path MTU and limiting TCP segments to this size, TCP can eliminate fragmentation at routers along the path that connect networks with different MTUs. Fragmentation adversely affects TCP throughput and network congestion.it is just requesting the packet size should be less than 1450, by default it will be 1500 in all the machine's. In case if you want reset to 1500, you can do it immediately with same command.