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

Re: [Condor-users] randomize_va_space - jobs evicted

On Fri, Feb 24, 2006 at 02:48:58PM +0100, Hermann Zellner wrote:
> Hello all,
> I am running condor-6.6.10 on Debian-Sarge with kernel and i686
> cpus. I notice again and again checkpoints which try to restart many
> times without success, but suddenly, after many hours, they restart
> normally.
> The reason for this is, as I strongly suspect, the stack-randomization,
> because by doing a
> echo 0 > /proc/sys/kernel/randomize_va_space
> on the executing host, I get such a lazy job to work again within five
> minutes.
> But this is no acceptable solution, because it disables a
> security-feature of the kernel. I even can't downgrade the kernel to a
> version lower than 2.6.12, when the stack-randomization was changed
> last.
> Is there a solution of this problem without disabling the
> stack-randomization? Has this problem been solved in condor-6.7.16 yet?
> But I would prefer to solve this problem within the stable 6.6.10,
> because everything else is working fine.

Yes, it works in 6.7. You can make it work in 6.6 by using a 
job wrapper script that does a 
setarch i386 ./my_job

at the end. 

This will disable the randomized address space for just the condor
job, and not the other processes.