Merge pull request #173 from EESSI/add_azure_CI_and_config

Add azure ci and config

CI/azure_mc/ci_config.sh

@@ -0,0 +1,4 @@
+# Configurable items
+if [ -z "${REFRAME_ARGS}" ]; then
+   REFRAME_ARGS="--tag CI --tag 1_node"
+fi

config/azure_mc.py

@@ -0,0 +1,86 @@
+# WARNING: for CPU autodetect to work correctly you need to
+# 1. Either use ReFrame >= 4.3.3 or temporarily change the 'launcher' for each partition to srun
+# 2. Either use ReFrame >= 4.3.3 or run from a clone of the ReFrame repository
+
+# Without this, the autodetect job fails because
+# 1. A missing mpirun command
+# 2. An incorrect directory structure is assumed when preparing the stagedir for the autodetect job
+
+# Related issues
+# 1. https://github.com/reframe-hpc/reframe/issues/2926
+# 2. https://github.com/reframe-hpc/reframe/issues/2914
+
+import os
+
+from eessi.testsuite.common_config import common_logging_config, common_general_config, common_eessi_init
+from eessi.testsuite.constants import FEATURES, SCALES
+
+# This config will write all staging, output and logging to subdirs under this prefix
+# Override with RFM_PREFIX environment variable
+reframe_prefix = os.path.join(os.environ['HOME'], 'reframe_runs')
+
+# AWS CITC site configuration
+site_configuration = {
+    'systems': [
+        {
+            'name': 'Magic_Castle_Azure',
+            'descr': 'Magic Castle build and test environment on Azure',
+            'modules_system': 'lmod',
+            'hostnames': ['login.*', '.*-node'],
+            'prefix': reframe_prefix,
+            'partitions': [
+                {
+                    'name': 'x86_64-amd-zen4-node',
+                    'access': ['--partition=x86-64-amd-zen4-node', '--export=NONE'],
+                    'descr': 'Zen4, 16 cores, 30 GB',
+                },
+                {
+                    'name': 'aarch64-neoverse-N1-16c-62gb',
+                    'access': ['--partition=aarch64-neoverse-n1-node', '--export=NONE'],
+                    'descr': 'Neoverse N1, 16 cores, 62 GiB',
+                },
+            ]
+        },
+    ],
+    'environments': [
+        {
+            'name': 'default',
+            'cc': 'cc',
+            'cxx': '',
+            'ftn': '',
+        },
+    ],
+    'logging': common_logging_config(reframe_prefix),
+    'general': [
+        {
+            # Enable automatic detection of CPU architecture for each partition
+            # See https://reframe-hpc.readthedocs.io/en/stable/configure.html#auto-detecting-processor-information
+            'remote_detect': True,
+            **common_general_config(reframe_prefix)
+        }
+    ],
+}
+
+# Add default things to each partition:
+partition_defaults = {
+    'scheduler': 'slurm',
+    'launcher': 'mpirun',
+    'environs': ['default'],
+    'features': [
+        FEATURES['CPU']
+    ] + list(SCALES.keys()),
+    'prepare_cmds': [
+        common_eessi_init(),
+        # Required when using srun as launcher with --export=NONE in partition access, in order to ensure job
+        # steps inherit environment. It doesn't hurt to define this even if srun is not used
+        'export SLURM_EXPORT_ENV=ALL'
+    ],
+    'extras': {
+        # Node types have strongly varying amounts of memory, but we'll make it easy on ourselves
+        # All should _at least_ have this amount
+        'mem_per_node': 64000
+    },
+}
+for system in site_configuration['systems']:
+    for partition in system['partitions']:
+        partition.update(partition_defaults)