Can't kill processes

[Cahyo Tri Nugroho]

Dear,

I have problem that I can't kill some processes in my system even as root. I've used 'kill -9' and forced kill from SAM but the process are still exist. Here are the process :

ediswit 9459 1 0 12:02:49 ? 0:00 mv ./appl/DOKRKSP.066249 ./appl/DOKRKSP.066252 ./appl/DOKRKS
ediswit 3731 1 0 11:35:00 ? 0:00 mv ./appl/DOKMANIF.066248 ./appl/DOKMANIF.066250 ./appl/DOKM
ediswit 9473 1 0 12:02:49 ? 0:00 mv ./appl/DOKRKSP.066249 ./appl/DOKRKSP.066252 ./appl/DOKRKS
ediswit 3733 1 0 11:35:00 ? 0:00 mv ./appl/DOKMANIF.066248 ./appl/DOKMANIF.066250 ./appl/DOKM
ediswit 9471 1 0 12:02:49 ? 0:00 mv ./appl/DOKRKSP.066249 ./appl/DOKRKSP.066252 ./appl/DOKRKS
ediswit 9464 1 0 12:02:49 ? 0:00 mv ./appl/DOKRKSP.066249 ./appl/DOKRKSP.066252 ./appl/DOKRKS
ediswit 3732 1 0 11:35:00 ? 0:00 mv ./appl/DOKMANIF.066248 ./appl/DOKMANIF.066250 ./appl/DOKM
ediswit 9468 1 0 12:02:49 ? 0:00 mv ./appl/DOKRKSP.066249 ./appl/DOKRKSP.066252 ./appl/DOKRKS
ediswit 23493 1 0 10:31:26 ? 0:00 mv ./appl/DOKRKSP.066249 ./appl/DOKRKSP.066252 ./appl/DOKRKS
ediswit 23483 1 0 10:31:26 ? 0:00 mv ./appl/DOKRKSP.066249 ./appl/DOKRKSP.066252 ./appl/DOKRKS
ediswit 9775 1 0 12:03:35 ? 0:00 mv ./appl/DOKRKSP.066249 ./appl/DOKRKSP.066252 ./appl/DOKRKS
ediswit 23502 1 0 10:31:26 ? 0:00 mv ./appl/DOKRKSP.066249 ./appl/DOKRKSP.066252 ./appl/DOKRKS
ediswit 9476 1 0 12:02:50 ? 0:00 mv ./appl/DOKRKSP.066249 ./appl/DOKRKSP.066252 ./appl/DOKRKS
ediswit 9462 1 0 12:02:49 ? 0:00 mv ./appl/DOKRKSP.066249 ./appl/DOKRKSP.066252 ./appl/DOKRKS
ediswit 23497 1 0 10:31:26 ? 0:00 mv ./appl/DOKRKSP.066249 ./appl/DOKRKSP.066252 ./appl/DOKRKS
ediswit 23489 1 0 10:31:26 ? 0:00 mv ./appl/DOKRKSP.066249 ./appl/DOKRKSP.066252 ./appl/DOKRKS

Please give me your suggestion how to kill them without system reboot :)

[john korterman]

Hi,

it is difficult to judge the reason without knowning more detailed circumstances, but your processes are probably waiting for some I/O to finish.
Check if you have a bad disk, a full filesystem, or perhaps just heavy I/O load.


regards,
John K.

[Peter Godron]

Hi,
can you tell us what your command was, as the mv command listed seems to be a bit confused over how many files it is supposed to move.
Normally you would have:
mv /path/a /path/b
You seem to have a lot of extra filenames

[Bill Hassell]

In Unix, the kill command only sets a signal. The -9 option sets a signal that a process can't ignore when it runs. Your problem is that these processes are waiting for an I/O task to complete and it hasn't, probably never will. About the only way this can happen is with networking, probably NFS or SAMBA shares where the remote machine is not working.

Because filesystems are expected to be working, there is no simple way to terminate the I/O. On the other hand, these processes aren't actually running so you don't need to worry about them. Work on the root cause of the problem (ie, networking?) and reboot later.

[Jean-Philippe HENRY_1]

Hello,

To supplement what Bill wrote, every process has its own signal table which is read only when goes from kernel mode (system calls) to user mode.

So if your process made an I/O in blocked mode it is still in kernel mode and cannot read its signal table. It happens mostly on devices that may be unplugged like network cables, USB keys, floppy disks, DVD, etc.

It is useless to send several times a same signal to a same process. For instance executing

kill -9 12345

only means to write a 1 in the 9th square of the signal table of the process whose ID is 12345. Sending another number 9 signal means to write a 1 in the same square where a 1 has already been written previously.

When the device where stands your filesystem becomes available, perhaps your process will end.

Regards,

JPH

[Jaydeb Chakraborty · Accepted solution]

First of all the mv command showing here is bit confused. However, if the file system is NFS mounted, it might be happened that NFS is down, hence the process is not being killed as its waiting for I/O.
Possible solution:
1) Use fuser to kill them
2) If NFS is the bottleneck, and it is down, then create a false IP-alias with same NFS IP in your working server and try to kill the process.

Thanks!

[Cahyo Tri Nugroho]

I'm sorry for very late reply. The problem was gone after I used following command :

fuser -ku nfsserver:/nfs_dir

Thank you for all of your suggestions.

[Cahyo Tri Nugroho]

I'm sorry for very late reply. The problem was gone after I used following command :

fuser -ku nfsserver:/nfs_dir

Thank you for all of your suggestions.