Help with error searching, suspected CDE problem

[Tommy Lindqvist]

Greetings exalted gurus :)

We are running HPUX11.00 on a 9000/785 machine.

We normally run remote from ReflectionX empowered PCs to connect by way of XDMCP.

For a long time now we had problems with the screen updating hanging for several seconds randomly about every 5-15 minutes.

First I suspected some network error like having switch at full duplex and nic at half, but after a couple of hours measuring throughput this was ruled out.

Second, I suspected incompatibilities with ReflectionX, but after experiencing same problems on console ( With CDE gui ) this also was ruled out.

System mounts home directories with automounter from SUN NFS server in case this can cause the symptoms.

I have installed the Mars clusterpack of patches.

I hope someone could tell me what tool I could use and what to measure to figure out what may be wrong, so the symptoms can go away.

Hope I am making sense.

Tommy

[RAC_1]

We will start with NFS mounted home dirs.

When screen update is taking place, do you get any nfs errors? Anything in syslog? dmesg? Also any errors in /var/dt/Xerrors and in $HOME/.t/errorlog?

Run dr_dt and check if you get any errors.

Anil

[Tommy Lindqvist]

/var/adm/syslog/syslog.log is (almost) empty, and no periodic errors. ( I had a couple of failed su's. )

$home/.dt/errorlog has no entries.

dmesg only messages regarding latest boot.

in /var/dt/Xerrors, I have a lot of complaints about my fontpath.

/usr/contrib/bin/X11/dr_dt finishes with no errors, and some 10 warnings about file permissions on links. ( And /dev/null )

Tommy

[Robert-Jan Goossens]

Hi Tommy,

-->
I have a lot of complaints about my fontpath
--<

out of curiosity, do you use your hpux server as a font server ?

[RAC_1]

When the screens hang, can you telnet to machine(or keep one ready beforehand) and do nfsstat (with -c and -s options) and check for bad calls, timeouts , retransmits.

Also make sure you have network settings correct. On sun NFS server, on clients(hp-ux, check as lanadmin -s lan_id) and on switches.

Anil

[Steve Steel]

Hi

Fix the /dev/null and any other warnings you can from dr_dt

see man xfs

start a font server

Which Xserver patch do you have


Steve STeel

[Tommy Lindqvist]

I will try and fix the rights.

I do run a fontserver.

I can run an xdmcp session from our PC to our SUN server, rlogin to our HP server, run a coupdl of graphical apps, and NOT see this problem. It is only when using the main CDE GUI directly either from console, or by choosing our HP server in the chooser from our pC clients.

These are a few of the patches installed.

PHSS_26620 B.11.00.12 Xserver cumulative patch
PHSS_28680 1.0 XClients Periodic Patch
PHSS_28869 1.0 ld(1) and linker tools cumulative patch
PHSS_29367 1.0 X/Motif 32bit Runtime Periodic Patch
PHSS_29369 1.0 X/Motif 64bit Runtime Periodic Patch
PHSS_29744 1.0 X Font Server Patch
PHSS_29963 1.0 HP DCE/9000 1.7 Runtime cumulative patch
PHSS_30010 1.0 CDE Runtime Patch

There are some 600 more patches ( From the QP-patch of Mars ) but these looked most relevant.

My thanks for the replies thus far, now somewhat rudely, I cannot answer any more replies for another sixteen hours.

Tommy

[Steve Steel]

Hi


Those are old patches

Example

Patch Name: PHSS_30181

Patch Description: s700_800 11.00 Xserver cumulative patch

Creation Date: 04/02/24

Post Date: 04/03/01

Hardware Platforms - OS Releases:
s700: 11.00
s800: 11.00

Products: N/A

Filesets:
Xserver.AGRM,fr=B.11.00,fa=HP-UX_B.11.00_32/64,v=HP
Xserver.DDX-ADVANCED,fr=B.11.00,fa=HP-UX_B.11.00_32/64,v=HP
Xserver.DDX-ENTRY,fr=B.11.00,fa=HP-UX_B.11.00_32/64,v=HP
Xserver.DDX-LOAD,fr=B.11.00,fa=HP-UX_B.11.00_32/64,v=HP
Xserver.DDX-SAM,fr=B.11.00,fa=HP-UX_B.11.00_32/64,v=HP
Xserver.DDX-SLS,fr=B.11.00,fa=HP-UX_B.11.00_32/64,v=HP
Xserver.DDX-UTILS,fr=B.11.00,fa=HP-UX_B.11.00_32/64,v=HP
Xserver.X11-SERV,fr=B.11.00,fa=HP-UX_B.11.00_32/64,v=HP
Xserver.X11-SERV-MAN,fr=B.11.00,fa=HP-UX_B.11.00_32/64,v=HP
Xserver.XEXT-DBE,fr=B.11.00,fa=HP-UX_B.11.00_32/64,v=HP
Xserver.XEXT-DBE-MAN,fr=B.11.00,fa=HP-UX_B.11.00_32/64,v=HP
Xserver.XEXT-DPMS,fr=B.11.00,fa=HP-UX_B.11.00_32/64,v=HP
Xserver.XEXT-DPMS-MAN,fr=B.11.00,fa=HP-UX_B.11.00_32/64,v=HP
Xserver.XEXT-HPCR,fr=B.11.00,fa=HP-UX_B.11.00_32/64,v=HP
Xserver.XEXT-HPCR-MAN,fr=B.11.00,fa=HP-UX_B.11.00_32/64,v=HP
Xserver.XEXT-MBX,fr=B.11.00,fa=HP-UX_B.11.00_32/64,v=HP
Xserver.XEXT-RECORD,fr=B.11.00,fa=HP-UX_B.11.00_32/64,v=HP

Automatic Reboot?: Yes

Status: General Release

Critical:
No (superseded patches were critical)
PHSS_29726: HANG PANIC
HANG - SR: 8606324111 CR: JAGae86576
PANIC - SR: 8606334950 CR: JAGae96027
PHSS_29182: ABORT HANG PANIC
PHSS_29168: ABORT


Get the latest patches


Steve Steel

[Mike Stroyan · Accepted solution]

Before you dig really deep into possible CDE problems you might confirm that it is just the user interaction that is hanging. There are patches such as PHKL_30190 that include fixes for symptoms that include brief system hangs. You could run

while date >> log; do sleep 1;done

Then go back and have a look at that log file to see if the script was able to run during a time when you saw a hang.

[Tommy Lindqvist]

Hi people,

Unfortunately I have been away on sick leave, so I havn't been able to try out the tests until now.

The
while ( 1 )
date; sleep 1
end

test produced the oddest result. Even during a 'freeze' when I couldn't change focus with the mouse, it continued to push out new times, punctually ever single second.

I will see if I can open up a parallell track with HP support. I had hoped it was something simple I had overlooked or missed from a FAQ. ( I have read them at least. Understand perhaps. )

Tommy

[Tommy Lindqvist]

Aha, when I modified the above, as per the suggestion, by pressing out the date >> logfile, I actually discovered a glitch.

The file wasn't written to during a 'freeze'

Thanks for that little gem. Network problems are somewhat easier to debug than CDE problems.

Time to trawl the patch database to see if the network patches are withing the QP - patchsets, or if they have to be downloaded separately.

I also need to check I run full duplex on both ends of the line.

Tommy