Can't log into B132L+

[Gary Cooper_1]

I've got a number of aging B132L+ workstations (that are soon to be replaced by B2600s). They are running HP-UX 11.11.

Each workstation has 3 NCD X-terminals connected to it, running a number of X-apps.

This morning, the X-terms lost communication with one of the workstations. I couldn't rlogin or telnet into it, but it responded to a ping. I went to the machine and the LEDs on the front panel looked OK, but the screen was just black - the usual CDE login screen wasn't being displayed.

Being unable to get in, I hit the power button, then rebooted it.

When it came back up, there were no suspicious messages in either syslog or my apllication log file.

The interesting thing is that the same thing happened to another two different workstations back in December.

Does anyone have any ideas what might cause these symptoms and what I might do to try and catch more information if it should happen again?

Thanks,

Gary

[Steve Steel]

Hi


Anything in /var/adm/syslog/OLDsyslog.log

Steve Steel

[Stf]

I think it's a resolving problem...
Verify your /etc/hosts,
Verify your /etc/nsswicth.conf if you have put DNS server on your machines...

Stf ;-)

[Gary Cooper_1]

Thanks for the replies guys.

No, I forgot to mention that OLDsyslog.log
was clear as well.

As far as the name resolution goes, this is a closed network that changes infrequently (although I've recently put a new Dell SBS 2003 machine and a couple of client PCs on the network).

nsswitch.conf looks like:
hosts: files [NOTFOUND=continue UNAVAIL=continue TRYAGAIN=continue] nis [NOTFOUND=continue UNAVAIL=continue TRYAGAIN=continue] dns[NOTFOUND=return UNAVAIL=return TRYAGAIN=return]
automount: nis [NOTFOUND=continue UNAVAIL=continue TRYAGAIN=continue] files

I must say that because the other two workstations exhibited the same problem at the same time, it did make me think NFS or autofs, but I couldn't see how that would stop the CDE login from being displayed.

Regards,

Gary

[Mohanasundaram_1]

Hi Gary,

Whenever I get into such problems or somebody says they are facing such problems, I always suggest to check if they have got a duplicate IP problem.

This is the most common reason, but very difficult to isolate. you need to format the /var/adm/nettl.LOG000 file using netfmt and view the formatted output to confirm this.

Did you say that more than 1 system experienced the same problem at the same time? then I would supsect the switch connecting them. Otherwise it can be the network interface or the LAN cables.

do let us know what you found. Also let me know if you need more help in formatting the nettl.LOG*** file.

If your name resolution, hostname,IP or network status changes then your CDE will exhibit such problems.

With regards,
Mohan.

[Pete Randall]

Gary,

Can you normally log into the workstations remotely via telnet or rlogin? (Other than when they're locked up, that is). If you normally can, then it sounds like a network connectivity issue - except you could still ping it. A lack of connectivity could account for CDE not working. In the CDE environment, name resolution and connectivity are critical. Check your LAN connection, the switch port, etc.

The problem is, if we can't get into the box when it's having the problem. It's going to be very hard to diagnose. How about setting up a cron job every hour to run a little health check script. Then you would have a log file you could check.


Pete

[Steve Steel]

Hi

Look at date of

/var/adm/nettl*

look at command

netfmt - format tracing and logging binary files


try
tail /etc/rc.log.old


Steve Steel

[Gary Cooper_1]

I've checked nettl.LOG000 and the timestamp on it is 12 Aug 2003 and doesn't contain anything interesting. Also there was nothing in rc.log.old.

I've also checked syslog on two of the other machines that it has NFS mounts onto and nothing.

On the switch/cable front, I would have thought that the fact that I could ping the machines would preclude this type of fault plus when this happened before, one X-term (that was already logged on to the workstation) continued to work OK although I couldn't get a CDE login screen up.

I had a look at the duplicate IP address thing and when I checked this afternoon, there weren't any duplicate IP addresses (I know this doesn't mean that there couldn't be, e.g if one of the machines was switched off when I did the check).

At the moment, the cron job to do a health check seems like the best option.

Regards,

Gary

[A. Clay Stephenson · Accepted solution]

When the only network related command that will work is "ping" that is a classic symptom that inetd has died. Look for and install any inetd related patches. As a matter of course on all of my boxes at least since HP-UX 9, I always have a cronjob that periodically checks to see if inetd is running and if not restart it. This has actually saved me a few times when otherwise I would have had to have someone reboot a remote machine or start inetd manually.

[Kent Ostby]

Gary -- if you get into this situation again, you should perform a Transfer of Control on the box so that a crash dump is created and can be looked at by HP.

To do this, you need a pen to press in the recessed TC button which is between the monitor port and Mouse/Keyboard ports on the B-Class.

After that is generated, HP can look at it and determine the cause.