[mesa-users] star/test fails in different ways depending on number of threads

Ehsan Moravveji e.moravveji at gmail.com
Fri Nov 22 00:46:22 EST 2013


Hi Tassos,
True. I got the same error recently, and managed to solve it by a trick.

I realized that every instance of MESA when launched requires almost 18GB of RAM at a very short time (maybe at 1milisec level) when it attacks for resources. Then, the required memory settles to ~1.2GB per each instance. 
Then, you have to see how much memory you have on each compute node, and manually specify the memory per each core.
Next, you must put a delay per each instance of MESA so that once the first manages to allocate its own resources, the second would wake up and so on.
For that, if you use SLURM, you can simply put a delay of 60 seconds (20 did not work for me) in your batch file like
sleep 60
However, if you use torque/PBS, it is more tricky. For that I use the rank number of the job (starting from 1) to and use the following simple formula in bash to put a pause:
sleep $(echo "60*($WORKER_RANK % 12)" | bc)
Since most of the machines here have 12 cores, I modulo the job rank (WORKER_RANK) with 12, and multiply that by 60 seconds, so I launch a new instance of MESA every minute.

If your job queueing system is different from SLURM and PBS, then you need to refer to the documentation.

I hope this helps.
Cheers.
Ehsan.




On Nov 21, 2013, at 11:48 PM, Aaron Dotter wrote:

> Hi Tassos,
> 
>                                          version_number        5596
>  Load1_eosDT_Table ierr        5014           5           4  0.80000000000000004       0.20000000000000001     
>  Load1_eosDT_Table ierr        5014           2           1  0.20000000000000001        0.0000000000000000     
> 
> This, specifically error code 5014, means a memory allocation error.  Ehsan recently reported a similar problem on a cluster.  Perhaps he has a solution.  Otherwise, see what you can about memory allocated to jobs you submit.
> 
> Aaron 
> ------------------------------------------------------------------------------
> Shape the Mobile Experience: Free Subscription
> Software experts and developers: Be at the forefront of tech innovation.
> Intel(R) Software Adrenaline delivers strategic insight and game-changing 
> conversations that shape the rapidly evolving mobile landscape. Sign up now. 
> http://pubads.g.doubleclick.net/gampad/clk?id=63431311&iu=/4140/ostg.clktrk_______________________________________________
> mesa-users mailing list
> mesa-users at lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/mesa-users

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.mesastar.org/pipermail/mesa-users/attachments/20131122/f30ac244/attachment.html>


More information about the Mesa-users mailing list