"ethtool -p" hanging at 100% CPU

[peetz]

Hi all,

I was just trying out the command

ethtool -p vmnic4[/code]

in one of our host's service console to visually identify one of the physical NICs (the command should make it blink its LEDs).

The command would hang consuming all of the COS' CPU cycles. I was not able to stop it, even "kill -9" from another root session did not work.

So I entered maintenance mode on the host, the VMs were all vmotioned away to other hosts, but now the task hangs at 37% at the "Disabling HA" stage.

Questions:

Has anyone used "ethtool -p" before with success? Would it work better with passing an additional argument telling the tool how long it should blink?

Any idea how I can successfully complete the "Enter maintenance mode" task?

\- Andreas

Message was edited by:

peetz

Update: The "Enter maintenance mode" task finally completed successfully (It just took very long), but now the machine hangs shutting down at "Stopping VMware ESX server services:"

[jonhutchings]

I don't know for sure but I thought ethtool -p caused ethtool to try to use a hardware specific feature of the nic card to flash the LEDs (typically). I suspect that since your are accessing the vmnic via the COS and not a physical device that the h/w feature is not implemented and accessing it has caused you to access a register/memory location that is being used by something else. It should really fail gracefully, but then maybe they didn't imagine anyone running code which tried to interface with the hardware in this manner.

This is just supposition on my part though