Deprecating 2022a modules on non-intel16 nodes

UPDATE: 2/24/2025 2:00PM - All affected modules have been removed from the main software library and are only available on intel16. Please follow the instructions in this post for managing this transition.

ICER will be making a set of modules unavailable on all clusters other than intel16. See this message for additional information.

What is happening?

ICER will be making a set of modules unavailable on all clusters other than intel16. The affected modules are any using the 2022a toolchain. This includes:

Any module with foss-2022a in the name
Any module with gompi-2022a in the name
Any module with GCC-11.3.0 in the name
Any module with GCCcore-11.3.0 in the name

For a full list, see the end of this announcement.

When is this happening?

If you use any of the affected modules, you will begin seeing a warning message immediately. These modules will be removed from non-intel16 nodes when the new cluster becomes publicly available on February 24th, 2025.

How does this affect me?

If you do not use any of the deprecated modules, you are unaffected and can continue as normal.

If you do use any of these modules, you have two options (or a hybrid of both):

Switch to a newer version

The first option is to use a newer version of these modules from a different toolchain. You can identify these newer versions by searching the module system either on the command line or in our documentation.

In some cases, you will need to restrict your jobs to only use certain node-types, since the newer modules may not be available everywhere. Check the specific software page linked in the overview in the documentation for details.

Example

Suppose that I currently use the module Amber/22.4-foss-2022a-AmberTools-22.5-CUDA-11.7.0 and want to use a newer version. I can log into a non-intel16 development node and search for newer versions:

$ ssh dev-amd20-v100
$ module avail Amber

------------------------------------ /opt/software-current/2023.06/x86_64/intel/skylake_avx512/modules/all -------------------------------------
   Amber/22.5-foss-2023a-AmberTools-23.6-CUDA-12.1.1 (D)

------------------------------------------- /opt/software-current/2023.06/x86_64/generic/modules/all -------------------------------------------
   Amber/22.4-foss-2022a-AmberTools-22.5-CUDA-11.7.0

This tells me that I can replace Amber/22.4-foss-2022a-AmberTools-22.5-CUDA-11.7.0 with Amber/22.5-foss-2023a-AmberTools-23.6-CUDA-12.1.1. I can also see this in the documentation page for Amber.

This page also tells me that the new version is available on amd22, amd24, intel18, amd21, and intel21 (even though amd20-v100 is listed, it refers to the development node, not the cluster type). So I should add the line

#SBATCH --constraint=[amd22|amd24|intel18|amd21|intel21]

to my job scripts.

Continue using old modules temporarily on `intel16` nodes only

These modules will remain on intel16. You can use the dev-intel16 development node, and constrain all jobs that need these modules to run on intel16 nodes only. To do so, add the following line to your SLURM script:

#SBATCH --constraint=intel16

Please note that these modules will only be available for as long as the intel16 cluster is available, which is scheduled to be retired before the start of the fall 2025 semester.

Both

You can combine the above techniques to continue using the old modules on intel16 nodes, but use the new modules on other nodes. This increases the range of nodes where your code can run.

Example

Using the Amber example from above, I can replace the section where I load modules in my SLURM script with

module purge

if [ "$HPCC_CLUSTER_FLAVOR" = "intel16" ]; then
    module load Amber/22.4-foss-2022a-AmberTools-22.5-CUDA-11.7.0
else
    module load Amber/22.5-foss-2023a-AmberTools-23.6-CUDA-12.1.1
fi

I should also use the constraint

#SBATCH --constraint=[intel16|amd22|amd24|intel18|amd21|intel21]

since these are the only cluster types where the Amber module (either version) is available.

Why is this happening?

These modules needed to use an older toolchain to be compatible with the older version of CUDA (11.7.0) required for the intel14 and intel16 GPUs. However, their toolchains are too old to be properly supported on modern hardware (like the new amd24 cluster).

To ensure that all newer hardware works properly, we are moving these modules to be available only on the intel16 cluster where there are no other alternatives. Since the intel16 cluster will be retired soon, these modules should be avoided anyways.

Full list of deprecated modules

Amber/22.4-foss-2022a-AmberTools-22.5-CUDA-11.7.0
APR/1.7.0-GCCcore-11.3.0
APR-util/1.6.1-GCCcore-11.3.0
Autoconf/2.71-GCCcore-11.3.0
Automake/1.16.5-GCCcore-11.3.0
Autotools/20220317-GCCcore-11.3.0
Bazel/5.1.1-GCCcore-11.3.0
binutils/2.38-GCCcore-11.3.0
Biopython/1.79-foss-2022a
Bison/3.8.2-GCCcore-11.3.0
BLIS/0.9.0-GCC-11.3.0
Boost/1.79.0-GCC-11.3.0
Bowtie2/2.4.5-GCC-11.3.0
Brotli/1.0.9-GCCcore-11.3.0
bzip2/1.0.8-GCCcore-11.3.0
CD-HIT/4.8.1-GCC-11.3.0
CESM-deps/2-foss-2022a
CMake/3.23.1-GCCcore-11.3.0
CMake/3.24.3-GCCcore-11.3.0
cppy/1.2.1-GCCcore-11.3.0
cURL/7.83.0-GCCcore-11.3.0
DB/18.1.40-GCCcore-11.3.0
dill/0.3.6-GCCcore-11.3.0
double-conversion/3.2.0-GCCcore-11.3.0
Doxygen/1.9.4-GCCcore-11.3.0
Eigen/3.4.0-GCCcore-11.3.0
ESMF/8.3.0-foss-2022a
expat/2.4.8-GCCcore-11.3.0
expecttest/0.1.3-GCCcore-11.3.0
FFmpeg/4.4.2-GCCcore-11.3.0
FFTW/3.3.10-GCC-11.3.0
FFTW.MPI/3.3.10-gompi-2022a
flatbuffers/2.0.7-GCCcore-11.3.0
flatbuffers-python/2.0-GCCcore-11.3.0
flex/2.6.4-GCCcore-11.3.0
FlexiBLAS/3.2.0-GCC-11.3.0
fontconfig/2.14.0-GCCcore-11.3.0
foss/2022a
freetype/2.12.1-GCCcore-11.3.0
FriBidi/1.0.12-GCCcore-11.3.0
GCC/11.3.0
GCCcore/11.3.0
GDRCopy/2.3-GCCcore-11.3.0
gettext/0.21-GCCcore-11.3.0
giflib/5.2.1-GCCcore-11.3.0
git/2.36.0-GCCcore-11.3.0-nodocs
GMP/6.2.1-GCCcore-11.3.0
gompi/2022a
gperf/3.1-GCCcore-11.3.0
groff/1.22.4-GCCcore-11.3.0
gzip/1.12-GCCcore-11.3.0
h5py/3.7.0-foss-2022a
HDF5/1.12.2-gompi-2022a
help2man/1.49.2-GCCcore-11.3.0
HH-suite/3.3.0-gompi-2022a
HMMER/3.3.2-gompi-2022a
HTSlib/1.15.1-GCC-11.3.0
hwloc/2.7.1-GCCcore-11.3.0
hypothesis/6.46.7-GCCcore-11.3.0
ICU/71.1-GCCcore-11.3.0
intltool/0.51.0-GCCcore-11.3.0
jbigkit/2.1-GCCcore-11.3.0
JsonCpp/1.9.5-GCCcore-11.3.0
LAME/3.100-GCCcore-11.3.0
libarchive/3.6.1-GCCcore-11.3.0
libdeflate/1.10-GCCcore-11.3.0
libevent/2.1.12-GCCcore-11.3.0
libfabric/1.15.1-GCCcore-11.3.0
libffi/3.4.2-GCCcore-11.3.0
libiconv/1.17-GCCcore-11.3.0
libjpeg-turbo/2.1.3-GCCcore-11.3.0
libpciaccess/0.16-GCCcore-11.3.0
libpng/1.6.37-GCCcore-11.3.0
libreadline/8.1.2-GCCcore-11.3.0
LibTIFF/4.3.0-GCCcore-11.3.0
libtool/2.4.7-GCCcore-11.3.0
libxc/6.2.2-GCC-11.3.0
libxml2/2.9.13-GCCcore-11.3.0
libxslt/1.1.34-GCCcore-11.3.0
libyaml/0.2.5-GCCcore-11.3.0
LLVM/14.0.3-GCCcore-11.3.0
LMDB/0.9.29-GCCcore-11.3.0
lxml/4.9.1-GCCcore-11.3.0
lz4/1.9.3-GCCcore-11.3.0
M4/1.4.19-GCCcore-11.3.0
magma/2.6.2-foss-2022a-CUDA-11.7.0
make/4.3-GCCcore-11.3.0
matplotlib/3.5.2-foss-2022a
Meson/0.62.1-GCCcore-11.3.0
MPFR/4.1.0-GCCcore-11.3.0
NASM/2.15.05-GCCcore-11.3.0
NCCL/2.12.12-GCCcore-11.3.0-CUDA-11.7.0
ncurses/6.3-GCCcore-11.3.0
netCDF/4.9.0-gompi-2022a
netCDF-C++4/4.3.1-gompi-2022a
netCDF-Fortran/4.6.0-gompi-2022a
networkx/2.8.4-foss-2022a
Ninja/1.10.2-GCCcore-11.3.0
nsync/1.25.0-GCCcore-11.3.0
numactl/2.0.14-GCCcore-11.3.0
OpenBLAS/0.3.20-GCC-11.3.0
OpenMPI/4.1.4-GCC-11.3.0
OSU-Micro-Benchmarks/5.9-gompi-2022a
Perl/5.34.1-GCCcore-11.3.0
Pillow/9.1.1-GCCcore-11.3.0
pkgconf/1.8.0-GCCcore-11.3.0
pkgconfig/1.5.5-GCCcore-11.3.0-python
PMIx/4.1.2-GCCcore-11.3.0
PnetCDF/1.12.3-gompi-2022a
protobuf/3.19.4-GCCcore-11.3.0
protobuf-python/3.19.4-GCCcore-11.3.0
pybind11/2.9.2-GCCcore-11.3.0
pytest-rerunfailures/11.1-GCCcore-11.3.0
pytest-shard/0.1.2-GCCcore-11.3.0
Python/2.7.18-GCCcore-11.3.0-bare
Python/3.10.4-GCCcore-11.3.0-bare
Python/3.10.4-GCCcore-11.3.0
PyTorch/1.13.1-foss-2022a-CUDA-11.7.0
PyYAML/6.0-GCCcore-11.3.0
Qhull/2020.2-GCCcore-11.3.0
Rust/1.60.0-GCCcore-11.3.0
SAMtools/1.16.1-GCC-11.3.0
ScaLAPACK/2.2.0-gompi-2022a-fb
scikit-build/0.15.0-GCCcore-11.3.0
SciPy-bundle/2022.05-foss-2022a
SCons/4.4.0-GCCcore-11.3.0
Serf/1.3.9-GCCcore-11.3.0
snappy/1.1.9-GCCcore-11.3.0
SQLite/3.38.3-GCCcore-11.3.0
Subversion/1.14.2-GCCcore-11.3.0
Szip/2.1.1-GCCcore-11.3.0
Tcl/8.6.12-GCCcore-11.3.0
TensorRT/8.6.1-foss-2022a-CUDA-11.7.0
Tk/8.6.12-GCCcore-11.3.0
Tkinter/3.10.4-GCCcore-11.3.0
TopHat/2.1.2-GCC-11.3.0-Python-2.7.18
TransDecoder/5.5.0-GCC-11.3.0
UCC/1.0.0-GCCcore-11.3.0
UCC-CUDA/1.0.0-GCCcore-11.3.0-CUDA-11.7.0
UCX/1.12.1-GCCcore-11.3.0
UCX-CUDA/1.12.1-GCCcore-11.3.0-CUDA-11.7.0
UnZip/6.0-GCCcore-11.3.0
utf8proc/2.7.0-GCCcore-11.3.0
util-linux/2.38-GCCcore-11.3.0
X11/20220504-GCCcore-11.3.0
x264/20220620-GCCcore-11.3.0
x265/3.5-GCCcore-11.3.0
XML-LibXML/2.0207-GCCcore-11.3.0
xorg-macros/1.19.3-GCCcore-11.3.0
XZ/5.2.5-GCCcore-11.3.0
Yasm/1.3.0-GCCcore-11.3.0
Zip/3.0-GCCcore-11.3.0
zlib/1.2.12-GCCcore-11.3.0
zstd/1.5.2-GCCcore-11.3.0

What is happening?

When is this happening?

How does this affect me?

Switch to a newer version

Example

Continue using old modules temporarily on intel16 nodes only

Both

Example

Why is this happening?

Full list of deprecated modules

Continue using old modules temporarily on `intel16` nodes only