Hung commands can't kill processes

[Gregory Locke]

On an HP 9000 N4000-65, I issue a ioscan -fn and the process hangs. From another sesson, as root, I can't kill the process(es) even with a -9 option.

On a new Itanium rx6600 I have the same issue with the df command. Hangs and can't be killed.

Is there any way to actually kill these processes without having to reboot these systems? Anybody have any suggestions?

All help greatly appreciated.

[Sundar_7 · Accepted solution]

The processes that are hung on I/O cannot be killed, even with the -9 signal.

You probably have a bad failed drive.

Execute dmesg or check the syslog.log for possible indications of a hardware failure.

[A. Clay Stephenson]

Processes waitng on high-priority events like i/o aren't actually running so the signals you send, while dutifully delivered to the process table, never are acted upon. In some cases, you can terminate the i/o event with an even higher priority i/o event by removing the device (powering off a tape drive, removing a failed disk drive) but otherwise these processes will not terminate because they actually aren't running.

[Bill Hassell]

Almost guarenteed you have NFS (or CIFS) mounts from another server and that server is broken or you have a SAN connected and something is messed up with the pathing (someone is playing with zones? BCV's not split correctly?). The hangs that can't be killed are due to I/O that has started but will never finish. In either case, HP-UX cannot bypass the problem. Use bdf or df to see where the list stops. The next mount point is the bad disk. Unfortunately, you may not be able to unmount the problem disk. A reboot will probably fix the hang by failing to mount the disk at reboot time. Check syslog for details.

[Gregory Locke]

Reboot the system