System hangup

[Irek_1]

I have HP9000 L-series HPUX11.00. It works for 2-3 hours, and after stop responding. If i try to open a new telnet session, i have only
"welcome" without login: On the console is the same problem. System responds only for ping. After this i can only turn off power. I've checked hardware with stm, no errors.
Could someone help me?
Irek

[Helen French]

What will happen to already running sessions ? Will that hang too? Did you check the OLDsyslog.log file? Any errors there? Apply latest patches and check root mail too.

[Steven E. Protter]

Lots of possible causes.

Perhaps nproc is too low in the kernel

perhaps nfiles is too low in the kernel

perhaps there are not enough pty drivers for terminal sessions in the kernel.

More likely is a networking problem. Bad configuration in /etc/rc.config.d/netconf or two NIC cards being broght up on the same network. This is a no no in HP-UX.'

SEP

[Rajeev Shukla]

Hi,
Adding to above,
Patching could be an issue, check that you have latest support plus installed.
Other major issue which could cause is NFS,do you have any NFS mounts, automounter. If the NFS performance patches are not installed it could make the machine run like a dog.

Cheers
Rajeev

[James A. Donovan]

Check the /var/adm/syslog/OLDsyslog.log file after having had to reboot from one of these hangups. It will probably tell you what's happening.

[ross martin_2]

Sounds like your pty pool has run out of available pty's and they are either all in use or the pty pool is not reclaiming the logged off pty's.

1. Increase the kernel parameter "npty" in SAM under Kernel Configuration
and then under Configurable Parameters

2. Regen the kernel and reboot with the new kernel

3. cd /dev

4. insf -d pty -n

5. insf -d ptym -n

6. cd /dev/pty

7. ln * ..

8. cd /dev/ptym

9. ln * ..

Hope this helps,

Ross

[Bill Hassell]

What does syslog say? Many sysadmins do not have any 'real' consoles where lots of useful messages are being written. I suspect you may find messages from the kernel complaining about running out of file handles or process table entries, etc.

[Irek_1]

I've checked OLDsyslog. In atachment is complete last OLDsyslog. It seem like NFS problems, i've just unmounted all nfs-directories, and i'll see. Maybe this is not the only problem and i schould do something else? This log ends at 00:36, and system had next hangup at 05:19 and nothing in syslog or mail! There was no users in the system and no users processes.
Thanks a lot!

[Brian Lee_4]

I had the same problem but I fixed it with changing Kernel parameters regarding maximun processes.

maxuprc : maximum processes per user
nprc : maximum processes for the system.

I increased the value of both kernel paramters and then ths system worked fine.

[Bill Hassell · Accepted solution]

Defintely sounds like NFS problems. NFS is a 'real' filesystem to HP-UX so when the server or network is clobbered, your system will hang. You'll see this during login, typing bdf and many other commands. HP-UX cannot tolerate filesystems just 'disappearing', just like it can't tolerate hard disks disappearing. Your NFS server(s)(and the network) must be much more reliable if they are being used on production machines.

[Chris Watkins_1]

Had something similar happen to my J5600 workstation last year.
(And on my manager's B2600 earlier this year)
(Not the nfs problems... but the hanging on telnet/ftp, and responding only to pings)
Mine turned out to be a disk going south... that wasn't picked up by diagnostics, either.

Replaced the disk, reinstalled, all is well.
What matches is that you're seeing nothing in mail or logs around the "hang" time.
The disk couldn't be written to, in our cases either, so nothing would get logged.

As for nfs mounts and their issues... whenever I'm forced (and I must be forced)
to use nfs mounts, I typically mount them with the "-o soft" option.
Keeps you from having to reboot a machine because the server holding one of its
filesystems went down for whatever reason.