I cant see vasp running in top command

Message

priyashrv · #1 Post by **priyashrv** » Wed Jul 24, 2013 12:55 pm

I try to run vasp5 on backend but when i do top it doesnt show vasp in there so i tried running in front end but my process stops with this error
For optimal performance we recommend that you set |
| NPAR = 4 - approx SQRT( number of cores) |
| (number of cores/NPAR must be integer) |
| This setting will greatly improve the performance of VASP for DFT. |
| The default NPAR=number of cores might be grossly inefficient |
| on modern multi-core architectures or massively parallel machines. |
| Do your own testing. |
| Unfortunately you need to use the default for hybrid, GW and RPA |
| calculations.

I am doing example 2_1_fccSi 2_2_fccSi_dos given in tutorials

please help !! i am newbie..pls ignore wrong usage /incorrect description

thank you

priyashrv · #2 Post by **priyashrv** » Wed Jul 24, 2013 1:20 pm

i m using ssh and command i gave is nohup mpirun -np 2 vasp5

#3 Post by **admin** » Thu Jul 25, 2013 12:25 pm

the 'error' message you actually show is not an error message, but a
warning (and should not be the reason why VASP stops).
please check
-- if you used the parallel executable
-- any possible output in one of the job's .o, .e, .po and .pe files
you have to login to one of the nodes on which vasp actually runs to see the process
via 'top'. However, if the output you show below is the last output written to OUTCAR, the executable crashes almost immediately, so you may simply not find the process within the time interval in which top updates its output.

btw:
these examples should not lead to any crash because of insufficient memory or stack size

priyashrv · #4 Post by **priyashrv** » Fri Jul 26, 2013 9:10 am

Yeah intially i had only 4 files but now i have so many files so vasp was actually running(i think so) but how do i know when it will finish if i cant see it running in top??

#5 Post by **admin** » Thu Sep 05, 2013 1:41 pm

you have to login on the MASTER NODE (and probably all other nodes) and kill the jobs explicitely. you WILL see them with top if you have a look on the correct nodes.