[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

re: [Condor-users] USB devices crisis...

Angel de Vicente <angelv@xxxxxx> wrote:
> > The information at http://www.cs.wisc.edu/condor/kernel.patch.html doesn't look
> > very clear to me:
> > 
> > "While Condor version 6.6.8 and above is able to detect keyboard and mouse
> > activity with the above Linux kernel revisions, it is only able to do so if you
> > are using a PS/2-style keyboard or mouse."
> > 
>Or does it mean that with version 6.6.7 we are OK because USB keyboard and mouse
>are detected? Actually I just opened a session in a Claimed/Busy machine with
>USB keyboard at our pool and the KeyboardIdle variable got updated correctly and
>eventually the Condor job was preempted. 

Did you open this session on the console keyboard, or remotely over the network via ssh etc ?   The issues discussed on the web page only relate to physically local keyboard detection...  remote sessions are still detected.

Here is the deal as I understand it :

Detection of ptty activity, such as happens via remote sessions (i.e. ssh) or even local xterm sessions, always works.  It does not matter if a USB keyboard was used or not.

Detection of physically local keyboard/mouse activity works fine with any Linux kernel provided you are (a) using a recent version of Condor and (b) using PS/2 style keyboards and mice.

IF you are using USB keyboards or mice, and you are using a 2.6 Linux kernel or a kernel that has some 2.6 changes back-ported into a 2.4 kernel (as redhat is sometimes known to do), detecting activity of your physically attached USB devices is a problem because the kernel functionality Condor relied upon to detect USB activity was removed in the 2.6 kernel.  Note removal of this functionality was supposedly done for kernel "efficiency", and has been hotly debated.    

You can solve the problem by installing the kernel patch referenced on the Condor web site.   

If you cannot install the patch, or downgrade the kernel, or purchase PS/2 keyboards to replace your USB keyboards :), then we need to get creative.

Do your users typically use X windows when sitting at their machines?  If so, perhaps we can come up with a solution.  The condor_kbdd daemon detects local keyboard activity by monitoring the X11 server, and it informs Condor of any activity it sees.  Typically we have used the condor_kbdd on platforms where Condor cannot see keyboard activity via the kernel such as on Compaq Tru64.   I have no idea if the condor_kbdd works on Linux or not, but if it does, it could solve the problem for folks using USB devices with the 2.6 Kernel who refuse to install the kernel patch.

What linux distro are you using?  Specifically, what kernel version and what X server?  (xfree86 or x.org version)


>Angel de Vicente
>PostDoc Software Support
>Instituto de Astrofisica de Canarias
>SIE Webpage
>Condor-users mailing list