ShutdownTimeout

[Paul_P_3]

Are there any negative ramifications of setting the ShutdownTimeout to 0 instead of a higher value such as 60 seconds?  The documentation states that

0. Setting value to zero indicates to simply reset the card without waiting for it to shut down.

I have a problem occasionally that in my TORQUE epilogue that sometimes a MIC fails to reset properly and while attempting to reset it the TORQUE epilogue times out and causes the epilogue to fail leaving the node in an unhealthy state.  Since the MICs memory is volatile and the OS gets reloaded at every boot, does it cause any issues to reset the OS on the MIC uncleanly?

Thanks,

Paul

[Loc_N_Intel]

The default shutdown timeout is set to 300 seconds. Although the micctrl utility allows users to set ShutdownTimeout to zero but it is not recommended since the coprocessor may enter an unknown state if doing so.

[Frances_R_Intel]

I don't know for sure what you are using in your script but if you are using 'micctrl -S -w', perhaps you could try doing the wait later in the epilogue:

do some work
micctrl -S
do some more work
micctrl -w

That will still guarantee that the coprocessors are shutdown and in the ready state before the script ends but I don't know if it is practical in your case or not. It depends on whether there is any other work to do.