[ncl-install] error while loading shared libraries: libgomp.so.1

[Dave Allured - NOAA Affiliate]

It is unclear when you are setting the environment variables.  I think you
need to set them at the start of each terminal session and each batch job.
If you did this only once when you installed the binary, then they would be
forgotten.  Try setting all three at the start of your batch job, before
your NCL command(s).

export NCARG_ROOT=...(My NCL path)
export PATH=$NCARG_ROOT/bin:$PATH
export LD_LIBRARY_PATH=/opt/gcc-4.6.1/lib:${LD_LIBRARY_PATH}

Watch out for quoting errors, and the third command may fail in some shells
if LD_LIBRARY_PATH is previously not defined.  If needed, use "env" to
check your settings.

I am really not sure what is going on.  If you are using module commands on
the supercomputer, they may interfere with these settings.  There is a
copacetic way to do this, and I do not have the local insight to offer
better advice.  You may also get better results if you use the most recent
NCL binary versions.  This version is several years old.

I strongly recommend that you show this problem to your local supercomputer
help desk, and ask for advice.  They are in charge of installing system
libraries, and they may easily be able to advise or fix this problem.


On Thu, May 14, 2020 at 5:13 AM Wen Yang via ncl-install <
ncl-install at ucar.edu> wrote:

> Thanks for your reply.
> I installed the NCL manually on the supercomputer.
> Here is my install procedure:
> I downloaded the "ncl_ncarg-6.4.0-RHEL6.4_64bit_gnu447.tar.gz" from the
> NCL website.
> Then extract it using the command "tar -xvf ..." and set the environment.
> "export NCARG_ROOT=...(My NCL path)
> export PATH=$NCARG_ROOT/bin:$PATH"
> Besides, when I use the command "which ncl_filedump", it shows the path
> where I installed NCL.
> I tried to run NCL again on my computer. I found that when I use the
> management nodes, I can run NCL without error.
> But if I submit the NCL by the compute nodes. It shows the error I just
> showed before.
>
> Thanks for your help.
>
> On Wed, May 13, 2020 at 3:25 PM Wen Yang <wyang123 at umd.edu> wrote:
>
>> Hi,
>> I was trying to run NCL by compute nodes. My system is gcc (GCC) 4.4.7
>> 20120313 (Red Hat 4.4.7-3).
>> But when I submit the job, it shows "*ncl: error while loading shared
>> libraries: libgomp.so.1: cannot open shared object file: No such file or
>> directory*".
>> Then I use the command "*locate libgomp.so.1* ", there is libgomp.so.1
>> in my linux system.
>> "*/opt/gcc-4.6.1/lib/libgomp.so.1*
>>
>>
>>
>>
>>
>> */opt/gcc-4.6.1/lib/libgomp.so.1.0.0/opt/gcc-4.6.1/lib64/libgomp.so.1/opt/gcc-4.6.1/lib64/libgomp.so.1.0.0/usr/lib64/libgomp.so.1*
>> */usr/lib64/libgomp.so.1.0.0"*
>> I add the path to the *LD_LIBRARY_PATH environment.*
>> "*export LD_LIBRARY_PATH=/opt/gcc-4.6.1/lib:${LD_LIBRARY_PATH}"*
>> But it doesn't work.
>> I am using the supercomputer, so I can't use the *sudo apt-get install*
>> command.
>> Could anyone help me with this?
>> Thanks,
>> Best,
>> Yang
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.ucar.edu/pipermail/ncl-install/attachments/20200517/a0a22d7f/attachment.html>