Hi Alan,

thanks for the explanations. I wasn't expecting the result of --module-only to be different depending on whether the actual installation of the software was already present or not. So that was the problem apparently. But you are right that it works properly when I do it again now.

I had to recreate the module because I forgot to change something in the easybuild configuration (module-depends-on=True). At the same time, I wanted to introduce architecture-specific builds for the different node types in our cluster that share the same module path on the login nodes. So it seems like the software was not built for this architecture yet, thus missing the installation dir and then the --module-only created modulefiles without the environment variables in them.

All solved now though, thanks again!

Best,
Maik


Am 11.06.2018 um 12:57 schrieb Alan O'Cais:
(EB version 3.6.0)

On Mon, 11 Jun 2018 at 12:56, Alan O'Cais 
<[email protected]<mailto:[email protected]>> wrote:
I just tested using '--module-only' on our (existing) installation and it set 
the envvars correctly. Are you sure you have the right idea about what 
--module-only does? If you use --module-only and --force on a blank slate you 
will get module files you describe since that skips the sanity check (which 
would've failed because the installation doesn't actually exist).

Alan

On Mon, 11 Jun 2018 at 12:37, Alan O'Cais 
<[email protected]<mailto:[email protected]>> wrote:
Hi Maik,

Can you explain that a little more? Why are you using '--module-only', are you 
re-using the CentOS 6 installation and creating a new module tree somewhere 
else? The guesses (I think) lead to the envvars being set in the module only if 
the something exists in the target, maybe send a debug log to see if these 
guesses are getting things wrong. It'll also be interesting to see what is 
being checked in the sanity check step that this passes but still generates an 
empty module.

Alan

On Mon, 11 Jun 2018 at 12:11, Maik Schmidt 
<[email protected]<mailto:[email protected]>> wrote:
Hm, could this be related to having used --module-only? I found this:

https://github.com/easybuilders/easybuild-framework/issues/1877

but the environment variables in GCCcore are not mentioned there. So I
guess this should have worked?

Best,
Maik

Am 11.06.2018 um 10:41 schrieb Maik Schmidt:
Hi all,

we're finally upgrading from RHEL6 to 7 and so I wanted to do fresh
builds of most of our software using EasyBuild. Now, when trying to
use the TensorFlow that I have built using the foss-toolchain, I got
the error:

ImportError: /usr/lib64/libstdc++.so.6: version `GLIBCXX_3.4.21' not
found (required by
/sw/installed/TensorFlow/1.8.0-foss-2018a-Python-3.6.4/lib/python3.6/site-packages/tensorflow/python/_pywrap_tensorflow_internal.so)

Apparently, it still uses the libstdc++.so from /usr instead of the
one from the GCCcore module. I have looked at the module file, and to
my surprise, I did not see it setting any environment variables like
LD_LIBRARY_PATH at all.

This cannot be intentional, right? It does not even set things like
PATH or MANPATH, so a:

$ module load GCC
$ which gcc
/usr/bin/gcc

still points to /usr, too. I've seen that in the gcc easyblock, there
is the function "make_module_req_guess" that looks like it is supposed
to handle those environment variables, but apparently, thoses
"guesses" are not used anywhere. At least it doesn't end up in the
module file for me... What am I missing here? Is this a bug?

Best,

Maik

--
Maik Schmidt
HPC Services

Technische Universität Dresden
Zentrum für Informationsdienste und Hochleistungsrechnen (ZIH)
Willers-Bau A116
D-01062 Dresden
Telefon: +49 351 463-32836




--
Dr. Alan O'Cais
E-CAM Software Manager
Juelich Supercomputing Centre
Forschungszentrum Juelich GmbH
52425 Juelich, Germany

Phone: +49 2461 61 5213
Fax: +49 2461 61 6656
E-mail: [email protected]<mailto:[email protected]>
WWW:    http://www.fz-juelich.de/ias/jsc/EN


--
Dr. Alan O'Cais
E-CAM Software Manager
Juelich Supercomputing Centre
Forschungszentrum Juelich GmbH
52425 Juelich, Germany

Phone: +49 2461 61 5213
Fax: +49 2461 61 6656
E-mail: [email protected]<mailto:[email protected]>
WWW:    http://www.fz-juelich.de/ias/jsc/EN


--
Dr. Alan O'Cais
E-CAM Software Manager
Juelich Supercomputing Centre
Forschungszentrum Juelich GmbH
52425 Juelich, Germany

Phone: +49 2461 61 5213
Fax: +49 2461 61 6656
E-mail: [email protected]<mailto:[email protected]>
WWW:    http://www.fz-juelich.de/ias/jsc/EN


------------------------------------------------------------------------------------------------
------------------------------------------------------------------------------------------------
Forschungszentrum Juelich GmbH
52425 Juelich
Sitz der Gesellschaft: Juelich
Eingetragen im Handelsregister des Amtsgerichts Dueren Nr. HR B 3498
Vorsitzender des Aufsichtsrats: MinDir Dr. Karl Eugen Huthmacher
Geschaeftsfuehrung: Prof. Dr.-Ing. Wolfgang Marquardt (Vorsitzender),
Karsten Beneke (stellv. Vorsitzender), Prof. Dr.-Ing. Harald Bolt,
Prof. Dr. Sebastian M. Schmidt
------------------------------------------------------------------------------------------------
------------------------------------------------------------------------------------------------



--
Maik Schmidt
HPC Services

Technische Universität Dresden
Zentrum für Informationsdienste und Hochleistungsrechnen (ZIH)
Willers-Bau A116
D-01062 Dresden
Telefon: +49 351 463-32836


Attachment: smime.p7s
Description: S/MIME Cryptographic Signature

Reply via email to