CDE stopped working

[Bill Sadvary]

All of a sudden, CDE stopped working on one of our RP3440 servers, HP-UX 11.23. It was working fine at one point. No recent hardware or OS changes.

This connects to a KVM along with three other RP3440s, that which have CDE working. All running same OS and patches.

I'm seeing a blank screen on the monitor but the KVM sees an active connection. If I disconnect the cable from the KVM, I see the connection go inactive, so it appears that the KVM is seeing signals from the graphics card when it is connected.

It's acting as if I did a "dtlogin.rc stop". In fact, I did a dtlogin.rc stop on one of the good servers and the display then went blank but still showed an active connection, just like the server with the CDE problem. So it seems like cabling and such is OK.

When I do a "dtllogin.rc start" from a serial console, I get the usual output (like on the good servers) output:

Using device file /dev/crt

**********
Starting Desktop login on host1.mydomain.edu:0...

Wait for login etc,etc.
**********

(It's the same output as on a good server, with the appropriate hostname, of course)

But the CDE login screen never appears and there are two dtlogin processes running.

root 16882 16868 0 11:42:08 ? 0:00 /usr/dt/bin/dtlogin

root 16891 16882 0 11:42:18 ? 0:00 /usr/dt/bin/dtlogin

If I set the DISPLAY variable, again from the serial console, to host1.mydomain.edu:0.0 and try running xclock and nothing appears on the monitor and no error messages on the console. The xclock is acting like everything worked OK but it didn't.

I went through the troubleshooting steps seen in several similar posts and I didn't find any fixes for my problem..

- It's not a DNS or /etc/hosts problem. nslookup resolves name and IP just fine

- Not a NIS problem. NIS is not in /etc/nsswitch either as seen here..

hosts: files [NOTFOUND=continue UNAVAIL=continue] dns

- I ran the dr_dt command and it's output is identical to a good server.

- I'm not running the font server but neither are the other three systems.

- The "setmon -r" gives the same results as a good server, same resolution, etc.

- /etc/dt/config/Xservers are identical with...
* Local local@console /usr/bin/X11/X :0

- No errors in /var/dt/Xerrors

- No clues in /var/adm/syslog/syslog.log

- Doing a "xhost" shows my FQDN

I am seeing the below error in /.dt/errorlog from some troubleshooting two days ago. I'm not sure what I did to cause that, it's not the result of a dtlogin.rc stop/start however.

*** ERROR(4): dtsession: PID 3757: Mon Oct 15 13:32:50 2007
Connection to server lost - exiting.
*** [96]

I starting to wonder is my graphics card is the problem.

Did I miss anything?

-Bill

[Alex Glennie]

looks like you covered most things ....

try disconnecting monitor from KVM switch and plug directly into graphics card . .... result ?

are remote CDE connections working OK to problematic server ?

Do you share mouse and keyboard too between all 3 servers ?

could this be an Xserver issue ? what happens if X is restarted ? Does the monitor repond to a ?

fyi 2 dtlogins are expected 1 is parent 2nd is spawned to service local display and/or any new remote connections.

I suspect the switch at this stage

[Bill Sadvary]

All four servers go to the KVM which then connects to a rackmount display (AB243AZ), which is kinda like a laptop. So yes, all four use same key and mouse (actually ball).

I can't connect the display cable directly to the server because they are different cables. Coming from the display is the typical monitor cable and two smaller round connectors (not usb) for the mouse and kybd.

On the server side, it's connecting to a A6150B graphics card via just one cable. This one cable has all three, the monitor, mouse and kybd bundled in.

How do I do a remote CDE connection? I can, from the problem server, display X to my workstation (sam, for example).

When I do the , nothing changes on the monitor but on the serial console, I get...

The X-server can not be started on display host1.mydomain.edu:0...
For details look into the file Dtlogin.errorLogFile (default: /var/dt/Xerrors)

--
Suspending Desktop Login...
Press enter for a login prompt

--

Log in. Desktop login will resume shortly after you logout.

--

I logout of the serial connection and nothing changes on the monitor, still blank. But at least we now know that it is accepting keyboard input from the rackmount diaplay/kybd and KVM.

Here's the errors in /var/dt/Xerrors..


-Start-

connection to bnrdb2.dickinson.edu:0.0 broken (explicit kill or server shutdown).

Wed Oct 17 14:36:51 2007
error (pid 27672): Server for display bnrdb2.dickinson.edu:0 terminated unexpectedly 0
XIO: fatal IO error 232 (Connection reset by peer) on X server "bnrdb2.dickinson.edu:0.0"
after 2375 requests (2367 known processed) with 194 events remaining.

Wed Oct 17 14:36:52 2007
error (pid 5551): Fatal IO error 232 ($s)

Wed Oct 17 14:36:54 2007
error (pid 27672): Server for display bnrdb2.dickinson.edu:0 terminated unexpectedly 0
XIO: fatal IO error 232 (Connection reset by peer) on X server "bnrdb2.dickinson.edu:0.0"
after 2375 requests (2367 known processed) with 187 events remaining.

Wed Oct 17 14:36:58 2007
error (pid 27672): Server for display bnrdb2.dickinson.edu:0 terminated unexpectedly 0
XIO: fatal IO error 232 (Connection reset by peer) on X server "bnrdb2.dickinson.edu:0.0"
after 2375 requests (2367 known processed) with 188 events remaining.

Wed Oct 17 14:37:02 2007
error (pid 27672): Server for display bnrdb2.dickinson.edu:0 terminated unexpectedly 0
XIO: fatal IO error 232 (Connection reset by peer) on X server "bnrdb2.dickinson.edu:0.0"
after 2375 requests (2367 known processed) with 188 events remaining.

Wed Oct 17 14:37:04 2007
error (pid 27672): Server for display bnrdb2.dickinson.edu:0 terminated unexpectedly 0
XIO: fatal IO error 232 (Connection reset by peer) on X server "bnrdb2.dickinson.edu:0.0"
after 2375 requests (2367 known processed) with 188 events remaining.

Wed Oct 17 14:37:04 2007
error (pid 27672): Server for display bnrdb2.dickinson.edu:0 can't be started.
X connection to bnrdb2.dickinson.edu:0.0 broken (explicit kill or server shutdown).
X connection to bnrdb2.dickinson.edu:0.0 broken (explicit kill or server shutdown).

Wed Oct 17 14:42:11 2007
error (pid 27672): Server for display bnrdb2.dickinson.edu:0 terminated unexpectedly 0
XIO: fatal IO error 232 (Connection reset by peer) on X server "bnrdb2.dickinson.edu:0.0"
after 2375 requests (2367 known processed) with 188 events remaining.
X connection to bnrdb2.dickinson.edu:0.0 broken (explicit kill or server shutdown).
X connection to bnrdb2.dickinson.edu:0.0 broken (explicit kill or server shutdown).

-End of Errors-

-Bill

[skt_skt]

did u try "/usr/dt/bin/dtlogin -e"

[Alex Glennie]

Bill,

remote cde session : by which I mean using ReflectionsX, Exceed or similar application to display the dtgreet cde login panel to a remote PC or xterminal .... I think it will work.

Looking at the errors it looks like X is failing to start not CDE ....

can u drop your run-level to 2 then back to 3 or attempt to start just X on the local display ? ie /usr/bin/X11/X :0

note errors then, I think X lost it's connection and isn't recognising attached monitor probably due to kvm ... I have seen this in the past but newer switches I assumed didn't have this problem ... I maybe wrong on this last point.

[Bill Sadvary]

I thinks it's my graphics card.

I stoppped dtlogin and the screen went to a random colored pattern, like an ugly plaid shirt. This is the first time that I seen anything other than a blank screen.

Then I did a ..

/usr/dt/bin/dtlogin -error /tmp/dtlogin.log -debug 1

and I could see, on the serial console, the stuff starting up. I still had the random pattern screen but I noticed that if I move the mouse ball, I can see an image moving around accordingly. Also, if I hit , I see debugging output on the serial console. I was even able to login (blindly) and logout (that was a fun challenge). So CDE is running, but I just can't see it.

Do you guys agree, it's a hardware issue (graphics card)?

Thanks!
-Bill

[Alex Glennie]

honest answer : not sure but what you describe sounds very similar to just X starting by itself see early suggestion. The fact the mouse moves a cursor also confirms this. It wouldn't work imho if there was a h/w issue.

Could you try the earlier suggestion(s) and compare results .... ?

Are we 100% nothing has changed, have you checked if file systems not full or close to full.

[Bill Sadvary]

OK, I'm backing up and doing some of the things I missed from above.

1.) I did a blind login, and from the serial console I can see the Xsession start, then did a . The screen kinda flickered, the Xsession process died and I had to login again (blindly). So that didn't help any. Seems to be working normally, just that I can't see anything.

When I login, it's acting just like a CDE login. After logged in, I have to do a right mouse click, which normally pulls up the "Workspace Menu", then I arrow down five times to "Logout" and then press return twice to confirm. At that point I am logged out and the Xsession dies.

2.) I tried the "/usr/bin/X11/X :0" from above, and it gave the error...

Fatal server error:
Cannot establish any listening sockets - Make sure an X server isn't already running.

Now that I think about it, an X server was already running:

daemon 28137 29110 0 13:23:08 ? 0:00 /usr/bin/X11/X :0 - auth /var/dt/bnrdbAAAa29110

Should I "kill -9" the above and try it again. It must have restarted sometime today, it has today's timestamp on it. Probably when starting dtlogin. ??

3.) I conncected a PC monitor to the graphics card and I'm seeing the exact same thing, a kinda random pattern of square colors with one big block of about 1"x1" which is the cursor.

Then only thing I haven't tried is the remote CDE login. Is that still worth a try at this point? If so, how do I display it to my HP workstation?

Thanks for help!!

-Bill

[Bill Sadvary]

Opps. Forgot to answer the most recent questions. Nothing has changed hardware wise and all four servers have been running the same OS for months now. All connecting through the same KVM and rackmount display.

/, /usr, /var, /tmp, and /stand are all less than 20% full.

-Bill

[Alex Glennie]

drop the run-level to 2 and run X:0 will prevent the Xserver restarting (it's CDE that's calling it every few seconds from run-level 3) or even comment out the Xservers like with reference to the local console but i'm not 100 % certain this will take effect by just killing X fyi that's what the does.

Lastly your description doesn't match the Xserver - it should be a consistent grey weave and the cursor should appear as a set of cross hairs so you maybe right wrt the graphics card.

good luck

[Bill Sadvary]

Thanks, Alex and Santhosh.

I logged a call with HW Support.

I'll let you know how it turns out.

-Bill

[Bill Sadvary]

The problem has been resolved and the fix required a new video card.

The HP Field Engineer couldn't order the original card (A6150B) so he installed an ATI Radeon 7500 in its place.

Of course, nothing is ever easy, so this involved me deinstalling two patches, PHKL_34349 (Fire GL-UX cumulative fixes) and PHSS_32960 (Xserver cumulative patch), installing the card, then re-applying the patches as mentioned in the thread..

http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=1175141

Then, following the instructions in the install guide got me through the rest of the configuration.

http://docs.hp.com/en/AB551-90003/AB551-90003.pdf

Thanks for all of the posts.

-Bill Sadvary

[Bill Sadvary]

Closing.