some java process can't be kill -9

[YLTan]

I have some stubborn java process that can be kill even with 'kill -9' run by root. How do I check what i/o, memory space or connection that it is tied to? I have lsof installed in this system.

Other than rebooting the server, is there a way to find out what causes it to ignore -9 signal and refuse to stop.

[Matti_Kurkela · Accepted solution]

A process that ignores kill -9 is often executing a system call. For some reason, the system call does not ever get finished.

First, find the PID of the stubborn process.

Check that the process is not already dead. If it is marked in the "ps -fp " listing, the process is already dead but its parent process is not willing to receive notification of the process's death. If you stop the parent process, the dead child process should also get cleaned up.

If you have NFS mounted disks, the processes that are using those disks will typically be stuck if the network connection to the NFS server fails.

If you have GlancePlus installed, it has a "Process Wait States" display (key W) which might help. The "Process System Calls" display might also be useful.

If you don't have Glance, do a "lsof -p " to see all the files the process is accessing. Are any of these on a NFS mounted disk?

Do a "dmesg | tail -20" and see if you have SCSI resets or other error messages there. /var/adm/syslog/syslog.log may also have some messages. Those might mean that some hardware is either rudely disconnected or broken.

[Ramkumar Devanathan]

Good comprehensive answer Matti.

- ramd.