Re: [Condor-users] KeyBoard Idle Time infinite

[Matthew Farrellee <matt@xxxxxxxxxxx>]

Todd Tannenbaum wrote:
> Saurabh Agarwal wrote:
> > Hello everybody.
> >
> > I just checked my Condor StartLog and to my surprise I get this message 
> > there which assumes infinite idle time for my mouse/keyboard stating:
> > "2/11 07:36:56 Unable to calculate keyboard/mouse idle time due to them 
> > both being USB or not present, assuming infinite idle time for these 
> > devices."
> >
> > Does anyone have any idea, why is it so? and how can I configure my 
> > condor to take in to account this measure,. I am sure the keyboard and 
> > mouse are present and yes they are connected through USB, so how do I 
> > make those changes to get such information. I mean I would like to know 
> > the information when are the users coming back to the machine, and 
> > "KeyboardIdle" will be the best ClassAd to give me such information.
> > I have installed Condor v 6.8.6 on my 3 machines, with one being the 
> > Central Manager.
>
> Condor does not know when there is activity on a USB connected keyboard 
> because the Linux kernel will not tell it.
>
> IIRC, you have a couple options.  The most straight-forward is change 
> the kernel via a small patch as discussed here:
>    http://www.cs.wisc.edu/condor/kernel.patch.html
> Over time many folks have requested many times for this patch to be 
> adopted into the mainline kernel, but so far it has not happened.
>
> Another possibility is to use the condor_kbdd if you are always running 
> an X server on your machine(s).  The condor_kbdd monitors keyboard/mouse 
> activity by asking the X server instead of the kernel, and it fowards 
> along what it discovers to the condor_startd.
>
> Anyone aware of any alternatives beyond the two above?
>
> regards,
> Todd

There are possibilities by reading /proc/interrupts, but it is 
non-trivial and might turn out to be somewhat system dependent.

Best,


matt