processes won't die (even with kill -9)

[Donal Murray]

I tried killing some processes with kill -9 , but they persisted (even when logged in as root). I then initiated a shutdown of the server in the hope this would resolve the problem. However, the same processes caused the shutdown sequence to hang, there is no response from the console, which means access to the system is impossible.

Q1: Is there any way I can regain acess without powering down?
Q2: What causes a process to get into a state where it will not die?

Thanks,

Donal

[Paula J Frazer-Campbell]

Hi
If the system is hung and you cannot connect to it, power down, reboot, and it should do a core dump on the way up, let HP check that out to see what they can tell you.

Apparently failed kill commands could be caused by a highly loaded machine and whilst the command seems to have failed it is waiting for prcessor time.
PS Can you telnet or ping the machine in its hung state?

[Donal Murray]

Hi Paula,
I can ping it, but not telnet to it. I guess I'll just have to bite the bullet and power down.

Thanks for your prompt response.

Donal

[Paula J Frazer-Campbell]

Hi Donal

If you can ping it the shutdown sequence has not yet stopped the lan card. This info might be useful to HP.

[June Hou]

Donal, here's a hypothesis:

If a process goes to sleep while it it's HP-UX priority value is below a certain level (I think it's 154), then the kernel won't send the process any signals until it wakes up.

One thing that can cause a long sleep is when a process is waiting on I/O... for instance, an NFS volume which the process is trying to read from has disappeared.

If you can figure out what the sleeping processes could be waiting on, you might be able to give 'em what they want, and get them to wake up. Only then will they receive whatever signals you sent them.

In actuality, they will only get the first signal -- one will do it. =-) Anyways, I hope that helps you the next time around.

-cheers, Craig S. Bell

[Suhas_2]

Hi,
Pls check whether the process is a defunct one. If so, check for its parent process ID.
Then try using kill -17 .
Do you have informix installed on your server ?...because there we have faced such problems.
Hope this helps....
Suhas...

[Donal Murray]

Hi Craig/Suhas,
Thanks for the advice. It was an oracle database writer process, which wouldn't shutdown. I'll keep both your points in mind next time around.

Regards,

Donal

[Carlos Fernandez Riera]

It is a oracle database?

I have this problem too. There is a new patch for this problem. I dont know PATCH_ID , sorry.