The latter is the approach we take.

ml EasyBuild # load stock EB modules

then

ml EasyBuild-ada # to build on/for Ivy Bridge nodes

or

ml EasyBuild-ada-Westmere # to build on/for Westmere nodes

or

ml EasyBuild-curie # to build on P7 nodes (this has its own P7 compile of EB)

with each having its own EASYBUILD_* settings (primarily _INSTALLDIR,
_REPOSITORYPATH and ROBOTPATH).

jack




On 09/12/2016 12:56 PM, Martin wrote:
We are rolling out a central /etc/profile.d/... file that exports the relevant EASYBUILD_ variables. Out method of choice is currently puppet but we are thinking about taking the intermediate of building a SiteEB module that does this (so "ml load SiteEB" instead of "ml load EasyBuild").

That way we decouple the upstream installation from our own site settings

On Mon, Sep 12, 2016 at 7:19 PM Åke Sandgren <[email protected] <mailto:[email protected]>> wrote:

    Why not keep /etc/easybuild.d/ in sync on all systems?

    On 09/12/2016 06:19 PM, Joachim Hein wrote:
    > Hi,
    >
    > The EasyBuild configuration file lives on a local disc of our
    > frontend /etc/easybuild.d/config.cfg hence EB is broken on any other
    > part of the cluster (e.g. building on a compute node).  I am aware
    > of
    
http://easybuild.readthedocs.io/en/latest/Configuration.html#configuration-file-s
    
<https://urldefense.proofpoint.com/v2/url?u=http-3A__easybuild.readthedocs.io_en_latest_Configuration.html-23configuration-2Dfile-2Ds&d=CwMFaQ&c=ODFT-G5SujMiGrKuoJJjVg&r=yuyoBkmTkIQPbv1BTF9U27ww5Lm7GhsMmWcQG9gmjbA&m=Y_NNis7GVIp_e7hV3npDq4Lr6iwZa3SBudImyPdjryY&s=G8Mng7i3L_TiYsEhPXKwgzDfY4lOvOBNf2BD4-dJwAY&e=>
    and
    > currently $EASYBUILD_CONFIGFILES seems the way forward. However
    setting
    > that as a system default in every account seems polluting to me.   A
    > natural place seems the EasyBuild module and I could hack it
    into there.
    >  However with the frequent upgrades to EB, I would need to
    remember to
    > hack the newly builded module each time.
    >
    > Since many here are operating multimode clusters, this must be a
    > frequent issue.  How are others handling this?
    >
    > Thanks and best wishes
    >   Joachim
    >

    --
    Ake Sandgren, HPC2N, Umea University, S-90187 Umea, Sweden
    Internet: [email protected] <mailto:[email protected]>  Phone: +46
    90 7866134 Fax: +46 90-580 14
    Mobile: +46 70 7716134 WWW: http://www.hpc2n.umu.se
    
<https://urldefense.proofpoint.com/v2/url?u=http-3A__www.hpc2n.umu.se&d=CwMFaQ&c=ODFT-G5SujMiGrKuoJJjVg&r=yuyoBkmTkIQPbv1BTF9U27ww5Lm7GhsMmWcQG9gmjbA&m=Y_NNis7GVIp_e7hV3npDq4Lr6iwZa3SBudImyPdjryY&s=-Vt9JMuIFrztszzhOTIk2ZVo8DFkVrNhDX5owscvPwQ&e=>

--
--
http://www.xing.com/profile/Martin_Marcher
http://www.linkedin.com/in/martinmarcher
Mobil: +43 / 660 / 62 45 103
UID: ATU68801424

Reply via email to