Author Topic: troubles with freshly installed cluster  (Read 7177 times)


  • Full Member
  • ***
  • Posts: 71
  • Karma: +0/-0
troubles with freshly installed cluster
« on: March 04, 2014, 11:43:13 PM »
Dear all,

I am trying to run TURBOMOLE on my newly installed cluster with SGE on it. I do everything as I should, e.g. add following lines to bash.profile

export TURBODIR=/home/jbaltrus/turbolinux_65/TURBOMOLE
export PATH=$TURBODIR/scripts:$PATH
export PATH=$TURBODIR/bin/em64t-unknown-linux-gnu_mpi:$PATH

ulimit -s unlimited


export PATH

but still optimization aborts right away with bunch of lines which I can;'t interpret, it seems like it starts OK but where it fails is unclear. Perhaps anyone has ideas?



 \\Warning: no access to tty (Bad file descriptor).
Thus no job control in this shell.
/home/jbaltrus/turbolinux_65/TURBOMOLE/bin/em64t-unknown-linux-gnu_mpi/sdg: line 1: ../em64t-unknown-linux-gnu/sdg: No such file or directory
ridft program used is /home/jbaltrus/turbolinux_65/TURBOMOLE/bin/em64t-unknown-linux-gnu_mpi/ridft
rdgrad program used is /home/jbaltrus/turbolinux_65/TURBOMOLE/bin/em64t-unknown-linux-gnu_mpi/rdgrad
relax program used is /home/jbaltrus/turbolinux_65/TURBOMOLE/bin/em64t-unknown-linux-gnu_mpi/relax
frog program used is /home/jbaltrus/turbolinux_65/TURBOMOLE/bin/em64t-unknown-linux-gnu_mpi/frog
uff program used is /home/jbaltrus/turbolinux_65/TURBOMOLE/bin/em64t-unknown-linux-gnu_mpi/uff
convgrep will be taken out of the TURBODIR directory
define program used is /home/jbaltrus/turbolinux_65/TURBOMOLE/bin/em64t-unknown-linux-gnu_mpi/define
statpt program used is /home/jbaltrus/turbolinux_65/TURBOMOLE/bin/em64t-unknown-linux-gnu_mpi/statpt
aoforce program used is /home/jbaltrus/turbolinux_65/TURBOMOLE/bin/em64t-unknown-linux-gnu_mpi/aoforce
/home/jbaltrus/turbolinux_65/TURBOMOLE/bin/em64t-unknown-linux-gnu_mpi/ridft: line 1: ../../mpirun_scripts/ridft: No such file or directory
ERROR in dscf step
Please look at the output file job.last
  Possible reasons:
     - unfinished step before running energy calculation
              ->  call \'actual -r\' and restart job
     - missing input files like coord, basis, control, ...
              ->  rerun define   
     - missing auxiliary basis set if using RI
              ->  switch on RI in define
     - mismatch between occupation and provided MOs
              ->  rerun define and redo \'eht\' step
/home/jbaltrus/turbolinux_65/TURBOMOLE/bin/em64t-unknown-linux-gnu_mpi/aoforce: line 1: ../em64t-unknown-linux-gnu/aoforce: No such file or directory


  • Sr. Member
  • ****
  • Posts: 216
  • Karma: +1/-0
Re: troubles with freshly installed cluster
« Reply #1 on: March 05, 2014, 08:48:21 AM »
Hi Jonas,

It's best to use the 'sysname' script to set the PATH to Turbomole executables:

For example:
export PATH=$PATH:$TURBODIR/scripts
export PATH=$TURBODIR/bin/`sysname`:$PATH

In particular, it's best to avoid appending '_mpi', '_smp' or any other qualifiers manually. If you want to run MPI jobs, just set


and redo

export PATH=$TURBODIR/bin/`sysname`:$PATH
(here it's important to put the "new" path before $PATH)

in your job script. Getting all the parallel schemes to work smoothly can take some time (, but generally things should work fine as long as you make sure that your PATH is updated in your SGE job script. The sample PBS script in the manual also outlines nicely the main points ( Exporting the whole environment with -V is also important (the same switch should work for SGE).



  • Global Moderator
  • Sr. Member
  • *****
  • Posts: 492
  • Karma: +0/-0
Re: troubles with freshly installed cluster
« Reply #2 on: March 05, 2014, 10:28:23 AM »
Hello Jonas,

Code: [Select]
/home/jbaltrus/turbolinux_65/TURBOMOLE/bin/em64t-unknown-linux-gnu_mpi/sdg: line 1: ../em64t-unknown-linux-gnu/sdg: No such file or directory
the sdg program is just a symbolic link to the serial version:

lrwxrwxrwx 1 huniar users       30 2013-03-05 17:12  sdg -> ../em64t-unknown-linux-gnu/sdg

the error you get could stem from the fact that there are no serial binaries installed in /home/jbaltrus/turbolinux_65/TURBOMOLE/bin/em64t-unknown-linux-gnu/ - could that be? Or is there a problem with file permissions (e.g. if you unpack the official Turbomole as root or with sudo)?