[ncl-install] Issue submitting an NCL script to slurm with sbatch: missing libraries

[Mary Haley]

Hi Elena,

It sounds to me like your environment is not quite right.

As an example, when I run a batch job on our supers and I want to use the
GNU version of NCL, I have to make sure I'm in the GNU compiler environment
before I run NCL:

module load gnu
module load ncl

Maybe you need to do something similar?

In general, you shouldn't have to mess with environment variables like
LD_LIBRARY_PATH.

--Mary


On Mon, Jan 7, 2019 at 2:23 AM Elena Kropac via ncl-install <
ncl-install at ucar.edu> wrote:

> Dear NCL support,
>
> I have an issue submitting an NCL script to slurm with sbatch. I am
> working in the field of Climate Modeling at a university in Germany and
> they provide a computing cluster which is a Linux RedHat 4.8.5-36 with gcc
> 4.8.5 (see below for further system information). I need to run an NCL
> script which uses the ESMF_RegridWeightGen function and takes too long to
> be run from the comand line (it runs without errors but is killed
> automatically after some 8 hrs because of computational expense).
> Therefore, I want to submit the script to slurm using sbatch. However, when
> I do so, I get the error "ncl: error while loading shared ibraries:
> libgomp.so.1: cannot open shared object file: No such file or directory".
>
> I have searched for the "missing" library using yum provides
> "*/libgomp.so.1" and found libgomp.so.1 (libgomp-4.8.5-36.el7.x86_64) in
> /usr/lib64/. I added this path to my LD_LIBRARY_PATH which did, however,
> not have any effect - I keep getting the same error.
>
> I currently have a pre-compiled version of NCL 6.4.0 installed
> (ncl_ncarg-6.4.0-RHEL6.4_64bit_nodap_gnu447.tar.gz), but I have already
> tried other binaries and versions. When I use the
> ncl_ncarg-6.5.0-CentOS7.5_64bit_nodap_gnu485.tar.gz one, which should match
> my compiler version, I get the same error, but refering to the
> libgfortran.so.3 library.
>
> I guess I might have an older version of the gfortran compiler on my
> system than what NCL was compiled with, but I do not quite know what the
> problem is, especially as the script is running in comand line without
> throwing me an error message. The error only appears when I try to submit
> the script to slurm. Do you have an idea what the problem could be and how
> to solve it? As I am working on a externally administrated cluster, I
> cannot simply install any files with sudo. But if that is necessary to
> get the script running, I can ask the system administrator.
>
> Here is some further information about my system:
>
> which ncl                 -> /home/hpc/gwgk/gwgi006h/LIBRARIES/ncl-6.4.0/bin/ncl
> ncl -V                    -> 6.4.0
> env | grep NCARG          -> NCARG_ROOT=/home/hpc/gwgk/gwgi006h/LIBRARIES/ncl-6.4.0
>                              NCARG_LIB=/home/hpc/gwgk/gwgi006h/LIBRARIES/ncl-6.4.0/lib
> which gcc                 -> /usr/bin/gcc
> which gfortran            -> /usr/bin/gfortran
> gfortran --version        -> GNU Fortran (GCC) 4.8.5 20150623 (Red Hat 4.8.5-36)
> gcc --version             -> gcc (GCC) 4.8.5 20150623 (Red Hat 4.8.5-36)
>
>
> Best regards,
>
> Elena Kropac
> _______________________________________________
> ncl-install mailing list
> List instructions, subscriber options, unsubscribe:
> http://mailman.ucar.edu/mailman/listinfo/ncl-install
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.ucar.edu/pipermail/ncl-install/attachments/20190107/c59d14f6/attachment-0001.html>