e4s-project / e4s-project.github.io Goto Github PK
View Code? Open in Web Editor NEWGitHub website for E4S Project
Home Page: https://e4s-project.github.io
License: MIT License
GitHub website for E4S Project
Home Page: https://e4s-project.github.io
License: MIT License
We need a page on e4s-project.github.io that describes how a product can join E4S. The page should also have a Join button at the top of pages, perhaps between "Contact Us" and "FAQ" or perhaps we could refactor one of the top buttons (such as "Contact Us") to include a link to the Join page.
The Join page should have the following information:
Pre-requisite: Justification for being included. The product must have some value or strong potential value to the HPC community. In other words, it must make sense for the product to belong to E4S. Note: It probably makes sense to have this pre-requisite be a new E4S community policy. See issue #8.
Level 0: Be listed in the E4S Spack installation script and be buildable in all E4S target environments. The list of Spack recipes for E4S is here: https://github.com/UO-OACISS/e4s/tree/master/e4s-facility-environments/uoregon.
Level 1: Be present in the DocPortal: https://e4s-project.github.io/DocPortal.html. To achieve this, we need a URL to your main repo to add to this list: https://github.com/E4S-Project/E4S-Documenter/blob/master/data/e4s_products.yaml.
Level 3: Satisfy all E4S community policies: https://e4s-project.github.io/policies.html. Full satisfaction is not required at this time. We are still establishing a process. Note that derived requirements from E4S member packages are not required to satisfy community policies as long as they do destabilize E4S builds or portability.
Hi,
I see E4S is redistributing Intel OneAPI compilers in containers:
We would like to do the same (ACCESS-NRI/model-builder#7) but I'm having trouble interpreting the Intel EULA. Does it allow this?
Please consider adding E4S Github org in home page or in Download section in https://e4s-project.github.io/index.html it should be visible on the top somewhere.
There seems to be two orgs. https://github.com/UO-OACISS/e4s and https://github.com/E4S-Project
This is confusing to end-users why there are two orgs, please consider merging these two into one place to avoid confusion.
I see a link https://github.com/E4S-Project/E4S-Project.github.io/blob/master/download.html#L104 to this repo in the download.html. It would help to put this on top somewhere.
This page needs to be updated https://e4s-project.github.io/installation.html
It would be nice if the page was better formatted to suit code format instead of scrolling from left to right which is quite annoying.
Anyhow, I ran into issue when running
salloc -N 2 -q interactive -t 00:30:00 --image=exascaleproject/sdk:AHM19 -C haswell -L SCRATCH
srun -n 32 shifter -- /bin/bash -c 'unset CRAYPE_VERSION; . /etc/bashrc ; spack load trilinos hypre parmetis hdf5 metis openblas superlu zlib netcdf matio [email protected] scalapack suite-sparse tau; spack unload openmpi mpich; ./Zoltan'
The issue is Zoltan is not in my current directory, if this package is provided by image then the path should be updated. Based on my discussion with Sameer the image needs to be updated so all examples for docker, singularity, shifter and charlie cloud would need to be reflected using a new image.
Right now i am unable to locate Zoltan so the srun
command will fail
I am trying to use the e4s build cache with spack
:
> spack mirror add E4S https://cache.e4s.io
> spack buildcache keys -it [238/391]
==> Fetching https://cache.e4s.io/build_cache/_pgp/25645FA2B218FE55B4EF649E4345F04B40005581.pub
==> Fetching https://mirror.spack.io/bootstrap/github-actions/v0.4/build_cache/linux-centos7-x86_64-gcc-10.2.1-patchelf-0.16.1-p72zyan5wrzuabtmzq7isa5mzyh6ahdp.spec.json
==> Fetching https://mirror.spack.io/bootstrap/github-actions/v0.4/build_cache/linux-centos7-x86_64/gcc-10
.2.1/patchelf-0.16.1/linux-centos7-x86_64-gcc-10.2.1-patchelf-0.16.1-p72zyan5wrzuabtmzq7isa5mzyh6ahdp.spac
k
==> Installing "patchelf@=0.16.1%gcc@=10.2.1 ldflags="-static-libstdc++ -static-libgcc" build_system=auto
tools arch=linux-centos7-x86_64" from a buildcache
==> Fetching https://mirror.spack.io/bootstrap/github-actions/v0.4/build_cache/linux-centos7-x86_64-gcc-10
.2.1-clingo-bootstrap-spack-cizcjj3tx4irs3uzoktrgybq73sz545f.spec.json ==> Fetching https://mirror.spack.io/bootstrap/github-actions/v0.4/build_cache/linux-centos7-x86_64/gcc-10
.2.1/clingo-bootstrap-spack/linux-centos7-x86_64-gcc-10.2.1-clingo-bootstrap-spack-cizcjj3tx4irs3uzoktrgyb
q73sz545f.spack
==> Installing "clingo-bootstrap@=spack%gcc@=10.2.1~docs~ipo+python+static_libstdcpp build_type=Release arch=linux-centos7-x86_64" from a buildcache
gpg: key 4345F04B40005581: public key "University of Oregon - E4S" imported
gpg: Total number processed: 1
gpg: imported: 1
gpg: inserting ownertrust of 6
However, installing packages fails with:
> spack install [email protected] [email protected]
==> Warning: you need a newer Spack version to read the buildcache index for the following mirror: 'https:
//cache.e4s.io'. The expected DB version is '6', but '7' was found.
# spack proceeds to install packages from source
spack v0.20.2 (Arch Linux)
Hello,
there are packages built for several operting systems according to the website.
https://oaciss.uoregon.edu/e4s/inventory.html
https://oaciss.uoregon.edu/e4s/packages/GCC.html
However listing packages on spack cli, there are fewer architectures available (e.g. centos, rhel).
[a spack]$ spack mirror list
E4S https://cache.e4s.io (fetch)
E4S https://cache.e4s.io (push)
spack-public https://mirror.spack.io
[a spack]$ spack buildcache list --allarch|grep '\--------'
-- linux-ubuntu18.04-ppc64le / [email protected] ------------------------
-- linux-ubuntu20.04-aarch64 / [email protected] ------------------------
-- linux-ubuntu20.04-x86_64 / [email protected] -------------------------
-- linux-ubuntu20.04-x86_64 / [email protected] -------------------
In the downloads page the simg files can be pushed to sylabs cloud.
http://tau.uoregon.edu/ecp.simg
http://oaciss.uoregon.edu/e4s/images/e4s_ubuntu1804_gpu_ppc64le_1.1.simg
Just create an account and follow instructions https://sylabs.io/guides/3.5/user-guide/cloud_library.html#pushing-a-container
It may make your life easier if you consider hosting docs on https://readthedocs.org/ that way you dont have to write html pages. You can still write pages in RST or Markdown and readthedocs will host your site. Just create an account https://readthedocs.org/accounts/signup/ and you can see the readthedocs guide https://docs.readthedocs.io/en/stable/index.html
I have used Sphinx for building your docs, it fairly simple to setup and used by many open-source projects.
Is there a reason that most of the E4S containers at dockerhub have an empty description/overview?
Speaking as a user, when I visit https://hub.docker.com/u/ecpe4s and see 107 different multi-GB docker containers where most/all have empty descriptions, it's very intimidating to find what I need. It's not even clear which of these actually contain full/partial E4S installs and which only contain precursors that can be used to build/install E4S. Even given my knowledge of E4S, looking only at dockerhub I was unable to identify the "best" docker container to recommend for UPC++ users on a Linux laptop without GPU support.
If the "best" E4S containers for end users are really 50+GB, then I think it would be valuable to provide a rich description of the contents and any system requirements. This would give the user confidence he's downloading something useful and not just wasting his time/data.
It's also notable that the vast majority of those 107 docker containers have not been updated in over a year; I'm guessing these are effectively "dead" and just left around for legacy reasons? It might also be valuable to partition the dockerhub account into one that contains only the full-featured containers maintained with the latest E4S installed (ie the "product" intended for end-users), and a separate dockerhub account that contains "everything else".
E4S is a portfolio of open-source, reusable software products to support high-performance computing applications. While this is a broad scope, this description does require that a candidate product has certain attributes and intents. Justifying why a particular product belongs in E4S should be the first community policy to address.
No detailed thoughts here, but it would be a good idea to have something simple and common sense. Perhaps it would be best to have a policy that each product has a code of conduct and not specify exactly what it should be.
Please consider adding a link to e4s slack channel. @tgamblin created a #e4s channel in spack if that is the preferred communication method then let's stick to that with a Join link from spack. You can use this link https://spackpm.herokuapp.com/ to autojoin the slack channel
When executing the e4S image as a user on a cluster node that already has LMod modules defined, I encounter the following problems:
wget http://tau.uoregon.edu/ecp.simg
singularity exec ./ecp.simg /bin/bash --rcfile /etc/bashrc
produces
Lmod Warning: The environment MODULEPATH has been changed in unexpected ways.
Lmod is unable to use given MODULEPATH. It is using:
" ... (my usual MODULEPATH, omitted here) ...".
Please use "module use ..." to change MODULEPATH instead.
/usr/bin/lua: /usr/share/lmod/lmod/libexec/MT.lua:1333: table index is nil
[...]
What steps are needed eithere to avoid passing in the existing MODULEPATH and LMod setup, or to use them (whichever is preferred to work with the E4S image)?
A declarative, efficient, and flexible JavaScript library for building user interfaces.
๐ Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. ๐๐๐
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google โค๏ธ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.