Merge pull request #337 from pyiron/release_openmpi_constraint

Release openmpi constraint - Test with openmpi 5.0

Author: jan-janssen

.ci_support/environment-openmpi.yml

@@ -3,7 +3,7 @@ channels:
 dependencies:
 - python
 - numpy
-- openmpi =4.1.6
+- openmpi
 - cloudpickle =3.0.0
 - mpi4py =3.1.6
 - tqdm =4.66.4

.github/workflows/benchmark.yml

@@ -53,4 +53,4 @@ jobs:
  cat timing.log
  python -m unittest tests/benchmark/test_results.py
  env:
- OMPI_MCA_rmaps_base_oversubscribe: 'yes'
+ PRTE_MCA_rmaps_default_mapping_policy: ':oversubscribe'

.github/workflows/unittest-flux-mpich.yml

@@ -0,0 +1,40 @@
+name: Unittests-flux-mpich
+
+on:
+ push:
+ branches: [ main ]
+ pull_request:
+ branches: [ main ]
+
+jobs:
+ build:
+ runs-on: ubuntu-latest
+ steps:
+ - uses: actions/checkout@v2
+ - name: Extend environment
+ shell: bash -l {0}
+ timeout-minutes: 5
+ run: |
+ echo -e '- coverage\n- flux-core =0.59.0\n- versioneer =0.28'>> .ci_support/environment-mpich.yml
+ cat .ci_support/environment-mpich.yml
+ - uses: conda-incubator/setup-miniconda@v2.2.0
+ with:
+ python-version: '3.12'
+ mamba-version: "*"
+ channels: conda-forge
+ miniforge-variant: Mambaforge
+ channel-priority: strict
+ auto-update-conda: true
+ environment-file: .ci_support/environment-mpich.yml
+ - name: Test
+ shell: bash -l {0}
+ timeout-minutes: 5
+ run: |
+ pip install . --no-deps --no-build-isolation
+ python -m unittest discover tests
+ - name: Test Flux
+ shell: bash -l {0}
+ timeout-minutes: 5
+ run: >
+ flux start
+ python -m unittest tests/test_flux_executor.py tests/test_executor_backend_flux.py;

.github/workflows/unittest-flux-openmpi.yml

@@ -0,0 +1,45 @@
+name: Unittests-flux-openmpi
+
+on:
+ push:
+ branches: [ main ]
+ pull_request:
+ branches: [ main ]
+
+jobs:
+ build:
+ runs-on: ubuntu-latest
+ steps:
+ - uses: actions/checkout@v2
+ - name: Extend environment
+ shell: bash -l {0}
+ timeout-minutes: 5
+ run: |
+ echo -e '- coverage\n- flux-core =0.59.0\n- flux-pmix=0.5.0\n- versioneer =0.28' >> .ci_support/environment-openmpi.yml
+ cat .ci_support/environment-openmpi.yml
+ - uses: conda-incubator/setup-miniconda@v2.2.0
+ with:
+ python-version: '3.12'
+ mamba-version: "*"
+ channels: conda-forge
+ miniforge-variant: Mambaforge
+ channel-priority: strict
+ auto-update-conda: true
+ environment-file: .ci_support/environment-openmpi.yml
+ - name: Test
+ shell: bash -l {0}
+ timeout-minutes: 5
+ run: |
+ pip install . --no-deps --no-build-isolation
+ coverage run -a --omit="pympipool/_version.py,tests/*" -m unittest discover tests
+ - name: Test Flux with OpenMPI
+ shell: bash -l {0}
+ timeout-minutes: 5
+ run: >
+ flux start
+ coverage run -a --omit="pympipool/_version.py,tests/*" -m unittest tests/test_flux_executor.py tests/test_executor_backend_flux.py;
+ coverage xml
+ env:
+ PYMPIPOOL_PMIX: "pmix"
+ - name: Coveralls
+ uses: coverallsapp/github-action@v2

.github/workflows/unittest-flux.yml

@@ -3,10 +3,11 @@ channels:
 dependencies:
 - python
 - numpy
-- openmpi =4.1.6
+- openmpi
 - cloudpickle =3.0.0
 - mpi4py =3.1.6
 - tqdm =4.66.2
 - pyzmq =26.0.0
 - flux-core =0.59.0
+- flux-pmix =0.5.0
 - versioneer =0.28

binder/environment.yml

@@ -51,8 +51,13 @@ conda install -c conda-forge flux-core flux-sched mpich=>4 pympipool
 Flux is not limited to mpich / cray mpi, it can also be installed in compatibility with openmpi or intel mpi using the 
 openmpi package: 
 ```
-conda install -c conda-forge flux-core flux-sched openmpi pympipool
+conda install -c conda-forge flux-core flux-sched openmpi=4.1.6 pympipool
 ```
+For the version 5 of openmpi the backend changed to `pmix`, this requires the additional `flux-pmix` plugin:
+```
+conda install -c conda-forge flux-core flux-sched flux-pmix openmpi>=5 pympipool
+```
+In addition, the `pmi="pmix"` parameter has to be set for the `pympipool.Executor` to switch to `pmix` as backend.
 
 ## Test Flux Framework
 To validate the installation of flux and confirm the GPUs are correctly recognized, you can start a flux session on the 

docs/installation.md

@@ -48,6 +48,7 @@ class Executor:
  function.
  init_function (None): optional function to preset arguments for functions which are submitted later
  command_line_argument_lst (list): Additional command line arguments for the srun call (SLURM only)
+ pmi (str): PMI interface to use (OpenMPI v5 requires pmix) default is None (Flux only)
 
  Examples:
  ```
@@ -85,6 +86,7 @@ def __init__(
  block_allocation: bool = True,
  init_function: Optional[callable] = None,
  command_line_argument_lst: list[str] = [],
+ pmi: Optional[str] = None,
  disable_dependencies: bool = False,
  refresh_rate: float = 0.01,
  ):
@@ -106,6 +108,7 @@ def __new__(
  block_allocation: bool = False,
  init_function: Optional[callable] = None,
  command_line_argument_lst: list[str] = [],
+ pmi: Optional[str] = None,
  disable_dependencies: bool = False,
  refresh_rate: float = 0.01,
  ):
@@ -143,6 +146,7 @@ def __new__(
  of the individual function.
  init_function (None): optional function to preset arguments for functions which are submitted later
  command_line_argument_lst (list): Additional command line arguments for the srun call (SLURM only)
+ pmi (str): PMI interface to use (OpenMPI v5 requires pmix) default is None (Flux only)
  disable_dependencies (boolean): Disable resolving future objects during the submission.
  refresh_rate (float): Set the refresh rate in seconds, how frequently the input queue is checked.
 
@@ -162,6 +166,7 @@ def __new__(
  block_allocation=block_allocation,
  init_function=init_function,
  command_line_argument_lst=command_line_argument_lst,
+ pmi=pmi,
  refresh_rate=refresh_rate,
  )
  else:
@@ -180,4 +185,5 @@ def __new__(
  block_allocation=block_allocation,
  init_function=init_function,
  command_line_argument_lst=command_line_argument_lst,
+ pmi=pmi,
  )

notebooks/examples.ipynb

@@ -13,6 +13,7 @@
  check_oversubscribe,
  check_executor,
  check_init_function,
+ check_pmi,
  validate_backend,
  validate_number_of_cores,
 )
@@ -50,6 +51,7 @@ def create_executor(
  block_allocation: bool = False,
  init_function: Optional[callable] = None,
  command_line_argument_lst: list[str] = [],
+ pmi: Optional[str] = None,
 ):
  """
  Instead of returning a pympipool.Executor object this function returns either a pympipool.mpi.PyMPIExecutor,
@@ -83,13 +85,15 @@ def create_executor(
  of the individual function.
  init_function (None): optional function to preset arguments for functions which are submitted later
  command_line_argument_lst (list): Additional command line arguments for the srun call (SLURM only)
+ pmi (str): PMI interface to use (OpenMPI v5 requires pmix) default is None (Flux only)
 
  """
  max_cores = validate_number_of_cores(max_cores=max_cores, max_workers=max_workers)
  check_init_function(block_allocation=block_allocation, init_function=init_function)
  backend = validate_backend(
  backend=backend, flux_installed=flux_installed, slurm_installed=slurm_installed
  )
+ check_pmi(backend=backend, pmi=pmi)
  if backend == "flux":
  check_oversubscribe(oversubscribe=oversubscribe)
  check_command_line_argument_lst(
@@ -105,6 +109,7 @@ def create_executor(
  cwd=cwd,
  executor=executor,
  hostname_localhost=hostname_localhost,
+ pmi=pmi,
  )
  else:
  return PyFluxStepExecutor(
@@ -115,6 +120,7 @@ def create_executor(
  cwd=cwd,
  executor=executor,
  hostname_localhost=hostname_localhost,
+ pmi=pmi,
  )
  elif backend == "slurm":
  check_executor(executor=executor)