system hung

[Michael Murphy_2]

Hello - i have an issue where my hpux system seems to have hung coming down - some of the filesystem apparently were "in use" and would not release. Now the system is not responding to any login attempts or sudo to root - i guess since /etc is umounted. I have a remote console - i asssume i need to go into console mode and restart the server (RS?) - does that sound like the best way or is there another that would umount the remainging file systems before rebooting? Thanks...

[Adam W.]

get on console. Find out what level it is in. Check the RC logs and syslog (if possible) use ps to attempt to find what processes are in use still.

[Michael Murphy_2]

console is not longer responding to anything besides ctnl-b to console mode - can't run shell commands or become root

[Adam W.]

get on the console. Find out what level it is in. Check the RC logs and syslog (if possible) use ps to attempt to find what processes are in use still.

[Michael Mike Reaser]

When you get on the console, can you look at the console logs to see how far along in the shutdown it got before it hung?

To me, this sounds more like a disk drive might be going flaky, and the system is waiting for forever-and-ever for it to unmount. *Or*...there are some NFS-mounted file systems that are hung because the remote system is down and/or unreachable - but I'd expect each such NFS unmount to eventually time out.

[Adam W.]

ok so, you cannot give ANY commands? At all? hmm, well, I would tell you to try umount but, at this point your only option may be to use "rs". I hope a more experienced Guru helps, cause That was the only 3 things that came to mind.

[Michael Murphy_2]

i cannot do any shell commands. i think i may have figured this out - seems that RS is the "harshest" reset method and maybe i should do a TC instead - anyone know if the TC will do an unmount of remaining filesystem?

[Adam W.]

yes rs is VERY "harsh" but worse things could happen. I am not certain if the TC will unmount BUT the TC command will reset the server with a crash dump which you can analyse later for troubleshooting.

[Adam W.]

Michael,
Have you changed the I.P. address recently?

[Michael Mike Reaser]

Both RS and TC are hardware-level commands, and have nothing to do with the operating system. Whatever you do now, your file systems will **NOT** be closed "gracefully", and it's likely - not guaranteed, but likely - that your system will be doing some fsck's on the way back up.

TC does a "T"ransfer of "C"ontrol and forces the system to reboot. RS does a "R"e"S"et of the system, re-initializing all memory and then rebooting. In a nutshell, what a TC does is a subset of what an RS does - a TC is a reboot *WITHOUT* the reinitialization of main memory.

If you've got HP support, I'd recommend placing a priority call with them, specifying that your system hung during shutdown and asking how you should proceed. If you don't have HP support and/or can't take the time to play "phone tag" with 'em, try a TC and reboot.

If the system hangs on the way up or otherwise doesn't boot cleanly after the TC, either place that call to HP support or try an RS. If the system still hangs or otherwise doesn't/won't boot after an RS, your only choice should really be to contact HP.

[Andres_13]

All GSP commands are OS independent, neither RS nor TC will attempt to umount anything. You ahould log a call within HP to get support on this issue.

Regards

[Patrick Wallek]

A TC will create a crash dump as well. So if need be, you could have this analyzed by HP to determine what the problem is/was.

An RS will NOT create a crash dump.

[Dennis Handly]

If you are hung during the shutdown and you see the progress messages, you might be able to cause rc(1) to skip the current step by: control-|