HPUX stop process error

[FenchurchLow]

HI all experts,

I am using hpux 11. when i tried to launch console at GUI, i got this error:

An attempt to launch process on "server" failed.
To continue, you may need to stop unneeded process on this host

I am not able to telnet to the server also.
The connection closed immediately.

Please help.

I really appreciate your help.

[Joelmel Roche]

In this case, try increasing swap

[whiteknight]

Hi Sdlow,

please check the /var/adm/syslog/syslog.log and dmesg output.

I suspected the system resources is out.


WK

[FenchurchLow]

Hi there,

Anything i try to launch at CDE is giving me the same error.

I cannot reboot or do anything.

[Dennis Handly]

>hpux 11.

(This is not helpful. Did you mean 11.00?)

Have you tried a command line logon?

>I cannot reboot or do anything.

What did you change to break it?
Do you have some runaway user script that is eating up the swap space?

Can you boot in single user mode?

[FenchurchLow]

HPUX 11.11

I cannot launch the terminal to run any commands

From CDE, if i select any functions, it will return the same error message.

[Timothy P. Jackson]

do a bdf and make sure that you did not run out of space on your root directory, especially /tmp

Tim

[Timothy P. Jackson]

do a bdf and make sure that you did not run out of space on your root logical volume, also check you /tmp logical volume

Tim

[skt_skt]

thorugh CDE , on the login screen selct the "fail safe"mode that allows the login when the other login process fails(even when unix console login fails).

But dtlogin should be enbaled/working on the server.In a similar scenario i did it through my reflecetionX program

[skt_skt]

thorugh CDE , on the login screen selct the "fail safe"(if what i rember is correct) mode that allows the login when the other login process fails(even when unix console login fails).

But dtlogin should be enbaled/working on the server

[skt_skt]

use "failsafe" mode on CDE login screen; that should permit the login in such scenario.. but dtlogin should be enabled/working on the server

[Dennis Handly]

>Tim: do a bdf and make sure that you did not run out of space

This would have to be done in single user mode.

>Santhosh: use "failsafe" mode on CDE login screen

I would assume the command line login would require less resources but it doesn't hurt to try but SDLow may have to boot into single user mode and try that bdf, swapinfo -tam and other options to see what is hitting the limit.

He may have to start a demon to record this and catch the evil user that has some kind of infinite fork loops in some shell script?

I'd hate to think it is something simple like nproc?

[FenchurchLow]

hi all,

i have to manually off the server to restart it back.

i notice alot of processes from oracle called

anyone faced this issue before?

[whiteknight]

Sdlow,

when you see a defunct process,
Defunct or zombie processes are processes that are sleeping
uninterruptibly while holding resources. If they never wake-up
to take a signal, they can remain on the system until it is
rebooted, holding their resources. Sending signals to defunct
(ps(1) will report "Z" for process state) processes using kill(1)
has no effect.

This behavior is normal and expected. When a process dies, it
becomes defunct and it remains on the system (less its address space)
until its parent reaps it using wait(2) to collect its exit value and
execution statistics.

Sending signals to an already dead process cannot make it any deader.
The process will be removed from the system (and the ps(1) output)
when it is reaped, otherwise the defunct process will remain until a
reboot.

See kill(1) and wait(2) for further information.


If defunct process is Oracle process, please log case to Oracle for help.

[Dennis Handly]

>whiteknight: otherwise the defunct process will remain until a reboot.

You can also kill zombies by killing the zombie master. When you kill the master, init(1M) will reap the zombie.
http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=1145648

(South Park had an episode about this, Pink Eye. :-)

[FenchurchLow]

Appreciate your response.

I am looking for the root cause of the error i got after i reboot overnight.

Will this kind of zombie processes eating the memory?

Kindly share with me.

[whiteknight]

Hi Dennis,

Thanks, never think of zombie master :)

Is there a simple way to trace zombie master ?


WK

[Dennis Handly]

>Will this kind of zombie processes eating the memory?

No, not swapspace. But it will take up an entry in the process table, that's why I mentioned nproc.

>WK: Is there a simple way to trace zombie master?

Master == evil parent that doesn't reap dead childred. :-) Look at the PPID.

[whiteknight]

SD Low,

Back to your question, yes defunct process is eating memory.

Are you planning for a reboot ?


WK

[Dennis Handly]

>WK: re you planning for a reboot?

I assumed he already had to because he couldn't use the GUI on the console nor telnet?

But if the system is still up and has problems, there are all sorts of things he can to to see if hitting some kernel parm limit.

Obviously "swapinfo -tam".
"ps -e | wc" and compare vs nproc.
Just listing the current kmtune values.

[FenchurchLow]

I rebooted the server using on/off button everytime i got this error.
It happens everyday when i return to check on the server.

[whiteknight]

SDlow,

please don't use on/off buttom to reset the server. This is not a correct procedure.

What model of the server ? I would advise to log case to HP Support to seek advise on this.



WK

[Dennis Handly]

>WK: please don't use on/off button to reset the server.

Using control B then "rs" is better in that the disks don't spin down.
(And it keeps the electrons and holes from getting lazy and attaching to each other. ;-)

[whiteknight]

Dennis,

:-) You bet!

[FenchurchLow]

ctrl+b is not working. that's why i used on/off :(

server model is rp4550