ccCA calculation failures?

Clicked A Few Times

8:35:35 AM PST - Thu, Nov 8th 2012

Hi all

I'm new to NWChem, and I'm trying to use NWChem to perform some ccCA calculations. I've gotten a few small calculations to work, but they seem to fail for relatively small systems, e.g. C4H8O3. I am running the jobs on a supercomputer with 372 nodes, each node having 8 processors and 24 GB of ram. I'm running off of /scratch, and my disk quota is 4 TB.

This is one of the jobs that failed. Using PBS, I gave it 16 GB of ram, 8 processors on one node, and a walltime of 6 h. It failed at the ccCA: MP2 calculation with aug-cc-pVQZ basis step, with the main error "mp2_grad:insufficient resources". A chunk of the end of the file is appended at the bottom.

I've been monitoring my disk quota, because the CBS- methods in Gaussian are also very disk intensive. However, I have more than 2 TB of my 4 TB remaining throughout the calculation. Any suggestions? Thanks!

Available:

 local mem=  1.15D+07

global mem=  2.61D+07

local disk=  1.09D+07

  1 passes of  12:        6143042                     660658                 153188418.

  2 passes of   6:        5847593                     375619                  76684942.

  3 passes of   4:        5749110                     280606                  51183784.

  4 passes of   3:        5699869                     233099                  38433204.

  5 passes of   3:        5699869                     233099                  38433204.

  6 passes of   2:        5650627                     185593                  25682625.

  7 passes of   2:        5650627                     185593                  25682625.

  8 passes of   2:        5650627                     185593                  25682625.

  9 passes of   2:        5650627                     185593                  25682625.

 10 passes of   2:        5650627                     185593                  25682625.

 11 passes of   2:        5650627                     185593                  25682625.

 12 passes of   1:        5601386                     138086                  12932046.

------------------------------------------------------------------------

mp2_grad:insufficient resources        1

------------------------------------------------------------------------

------------------------------------------------------------------------

 current input line :

   19: task ccca

------------------------------------------------------------------------

------------------------------------------------------------------------

There is an error in the input file

------------------------------------------------------------------------

Forum Vet

10:11:44 AM PST - Thu, Nov 8th 2012
Andrew, What platform have you been running NWChem on and what version of NWChem? Thanks, Edo

Clicked A Few Times

11:56:10 AM PST - Thu, Nov 8th 2012
I'm using NWChem 6.1.1 on LINUX64. $ cat /proc/version Linux version 2.6.18-164.15.1.el5.perfctr (fdang@node314) (gcc version 4.1.2 20080704 (Red Hat 4.1.2-46)) #1 SMP Wed Feb 22 16:28:05 CST 2012 I'm wondering if my jobs are failing for insufficient memory. The defaults are 400 mb total. I've changed it to 4096 or 8192 mb for a 4-proc job, and it seems to be behaving a little better.

Forum Vet

12:46:00 PM PST - Thu, Nov 8th 2012
Quote:Andrew.yeung Nov 8th 10:56 am I'm wondering if my jobs are failing for insufficient memory. The defaults are 400 mb total. I've changed it to 4096 or 8192 mb for a 4-proc job, and it seems to be behaving a little better. Yes, if you give more memory you should be able to get the MP2 job running. Were you able to pass the previous point where the job stopped? Edo

Clicked A Few Times

12:50:07 PM PST - Thu, Nov 8th 2012
So far so good, thanks. Keeping my fingers crossed right now. Sorry if this is a silly question, I'm used to how G09 is set up here, and I didn't think that memory would have to be explicitly defined. Is there a way I can use my .job-creating script to set an environment variable that defines how much memory to use?

Forum >> NWChem's corner >> Running NWChem