Hey Guys

I have not run into this one before. I tried to open an unknown file with gedit. It hung (not so unusual), but then I tried to kill it and it will not die (unusual for me). So what does one do when 'kill -9 3144' (the pid) does not work? It takes the command without error but it does not kill it.

Thanks
Lazlow

1) CTRL+ALT+BACKSPACE

or

2) shutdown -r now

I am trying to avoid that. It may be several hours before I can shut the machine down and the process is taking 50% of the cpu.

maybe try killall

Nigel

No go, but thanks for the suggestion. I really cannot figure this one out. It is showing an RN (running) and not z (zombie) or d (I think of it as dead but not the right term). The little !@#$$%%^ is just sitting there chewing up cycles. It also does not show that it is a daughter process for anything.

Can you kill the xterm session you started it from?

No, I just double clicked on it(no xterm).

-11 will tell the kernel to exterminate with no pity!

The kernel thinks the process is trying to access mem addresses that are not assigned to the process and will terminate it immediately. gedit might dump core, but that is not really an issue ... out of curiosity, what were you trying to open, a very big file?

The other option would be to use xkill, it is also an applet for the gnome-panel and allows you to kill xwindow processes pretty darn well ... if that does not work, a reboot is required.

I take it logging off doesn't help

how about
top
r
PID to renice 19 then k

Nigel

No, logging off did nothing(multiple times). I even tried logging in as root (I know but I was desperate).

Tried to renice earlier(good idea though).

Yeah, I tried -11 (forget where I found that one) no luck. When the app still has a window on the desktop xkill works pretty good, unfortunately the window had been closed long before I realized gedit was hung (ok, really hung). Yes, it was a very big file. One of those times I just tried to open something before I noticed how big it was. I have hung gedit this way before but it has always died respectfully.

All right guys I can (and will) reboot the system now. If anybody has more ideas please go ahead and post them. The next time someone hangs a process, this thread should serve as a good reference.

Thanks for all the help.

Lazlow

If a kill -9 fails the process is probably a zombie.

http://en.wikipedia.org/wiki/Zombie_process

As it's taking CPU cycles, it could be IO bound and in this case it will not respond to a kill until it's been given th resources that it's requesting.

You could probably put it into a cgroup container and not give it (the container) any service. I'll test this later.

Dissillusionist

As I stated in post #5 there were no Zs (zombies) or Ds(dead processes).


sysg00

Not a clue what you are talking about(my ignorance showing again). Can you move an already running process like this?

Yep, that's the beauty of it - have a look at ../Documentation/cgroups.txt for the basic design.
Used to be (when it was called cpusets) you could only limit by (entire) CPU - (i.e. 1 out of 4 say), not a percentage like 15%. That's how I've used it, but I think CPU controller has been updated to do this (finer control) but I'll have to check. Might only be on (really) recent kernels tho'.

Testing may have to wait - real work is getting in the way ....