Merge branch 'charmed-hpc:main' into nuccitheboss/impl-sackd-manager

dev-requirements.txt

@@ -1,5 +1,5 @@
 # lib deps
-slurmutils ~= 0.8.0
+slurmutils ~= 0.9.0
 python-dotenv ~= 1.0.1
 pyyaml >= 6.0.2
 distro ~=1.9.0

lib/charms/hpc_libs/v0/slurm_ops.py

@@ -77,8 +77,8 @@ def _on_install(self, _) -> None:
 import yaml
 from cryptography.hazmat.primitives import serialization
 from cryptography.hazmat.primitives.asymmetric import rsa
-from slurmutils.editors import cgroupconfig, slurmconfig, slurmdbdconfig
-from slurmutils.models import CgroupConfig, SlurmConfig, SlurmdbdConfig
+from slurmutils.editors import acctgatherconfig, cgroupconfig, slurmconfig, slurmdbdconfig
+from slurmutils.models import AcctGatherConfig, CgroupConfig, SlurmConfig, SlurmdbdConfig
 
 try:
     import charms.operator_libs_linux.v0.apt as apt
@@ -97,14 +97,14 @@ def _on_install(self, _) -> None:
 
 # Increment this PATCH version before using `charmcraft publish-lib` or reset
 # to 0 if you are raising the major API version
-LIBPATCH = 8
+LIBPATCH = 9
 
 # Charm library dependencies to fetch during `charmcraft pack`.
 PYDEPS = [
     "cryptography~=43.0.1",
     "pyyaml>=6.0.2",
     "python-dotenv~=1.0.1",
-    "slurmutils~=0.8.3",
+    "slurmutils~=0.9.0",
     "distro~=1.9.0",
 ]
 
@@ -265,6 +265,28 @@ def edit(self) -> SlurmConfig:
             yield config
 
 
+class _AcctGatherConfigManager(_ConfigManager):
+    """Manage the `acct_gather.conf` configuration file."""
+
+    def load(self) -> AcctGatherConfig:
+        """Load the current `acct_gather.conf` configuration file."""
+        return acctgatherconfig.load(self._config_path)
+
+    def dump(self, config: AcctGatherConfig) -> None:
+        """Dump new configuration into `acct_gather.conf` configuration file."""
+        acctgatherconfig.dump(
+            config, self._config_path, mode=0o600, user=self._user, group=self._group
+        )
+
+    @contextmanager
+    def edit(self) -> AcctGatherConfig:
+        """Edit the current `acct_gather.conf` configuration file."""
+        with acctgatherconfig.edit(
+            self._config_path, mode=0o600, user=self._user, group=self._group
+        ) as config:
+            yield config
+
+
 class _CgroupConfigManager(_ConfigManager):
     """Control the `cgroup.conf` configuration file."""
 
@@ -921,6 +943,9 @@ def __init__(self, *args, **kwargs) -> None:
         self.config = _SlurmConfigManager(
             self._ops_manager.etc_path / "slurm.conf", self.user, self.group
         )
+        self.acct_gather = _AcctGatherConfigManager(
+            self._ops_manager.etc_path / "acct_gather.conf", self.user, self.group
+        )
         self.cgroup = _CgroupConfigManager(
             self._ops_manager.etc_path / "cgroup.conf", self.user, self.group
         )

tests/unit/test_slurm_ops.py

@@ -166,6 +166,27 @@
 PartitionName=DEFAULT MaxTime=30 MaxNodes=10 State=UP
 PartitionName=batch Nodes=juju-c9fc6f-2,juju-c9fc6f-3,juju-c9fc6f-4,juju-c9fc6f-5 MinNodes=4 MaxTime=120 AllowGroups=admin
 """
+EXAMPLE_ACCT_GATHER_CONFIG = """#
+# `acct_gather.conf` file generated at 2024-09-18 15:10:44.652017 by slurmutils.
+#
+EnergyIPMIFrequency=1
+EnergyIPMICalcAdjustment=yes
+EnergyIPMIPowerSensors=Node=16,19;Socket1=19,26;KNC=16,19
+EnergyIPMIUsername=testipmiusername
+EnergyIPMIPassword=testipmipassword
+EnergyIPMITimeout=10
+ProfileHDF5Dir=/mydir
+ProfileHDF5Default=ALL
+ProfileInfluxDBDatabase=acct_gather_db
+ProfileInfluxDBDefault=ALL
+ProfileInfluxDBHost=testhostname
+ProfileInfluxDBPass=testpassword
+ProfileInfluxDBRTPolicy=testpolicy
+ProfileInfluxDBUser=testuser
+ProfileInfluxDBTimeout=10
+InfinibandOFEDPort=0
+SysfsInterfaces=enp0s1
+"""
 EXAMPLE_CGROUP_CONFIG = """#
 # `cgroup.conf` file generated at 2024-09-18 15:10:44.652017 by slurmutils.
 #
@@ -653,6 +674,45 @@ def test_config(self, *_) -> None:
         self.assertEqual(f_info.st_gid, FAKE_GROUP_GID)
 
 
+class TestAcctGatherConfig(TestCase):
+    """Test the `slurmctld` service acct_gather configuration manager."""
+
+    def setUp(self) -> None:
+        self.setUpPyfakefs()
+        self.manager = SlurmctldManager(snap=True)
+        self.fs.create_file(
+            "/var/snap/slurm/common/etc/slurm/acct_gather.conf",
+            contents=EXAMPLE_ACCT_GATHER_CONFIG,
+        )
+
+    def test_config(self) -> None:
+        """Test that manager can manipulate cgroup.conf configuration file."""
+        # Fake user and group that owns `cgroup.conf`.
+        self.manager.acct_gather._user = FAKE_USER_NAME
+        self.manager.acct_gather._group = FAKE_GROUP_NAME
+
+        with self.manager.acct_gather.edit() as config:
+            self.assertEqual(config.energy_ipmi_frequency, "1")
+            self.assertEqual(config.energy_ipmi_calc_adjustment, "yes")
+            self.assertListEqual(config.sysfs_interfaces, ["enp0s1"])
+
+            config.energy_ipmi_frequency = "2"
+            config.energy_ipmi_calc_adjustment = "no"
+            config.sysfs_interfaces = ["enp0s2"]
+
+        # Exit the context to save changes to the file
+        config = self.manager.acct_gather.load()
+        self.assertEqual(config.energy_ipmi_frequency, "2")
+        self.assertEqual(config.energy_ipmi_calc_adjustment, "no")
+        self.assertListEqual(config.sysfs_interfaces, ["enp0s2"])
+
+        # Ensure that permissions on file are correct.
+        f_info = Path("/var/snap/slurm/common/etc/slurm/acct_gather.conf").stat()
+        self.assertEqual(stat.filemode(f_info.st_mode), "-rw-------")
+        self.assertEqual(f_info.st_uid, FAKE_USER_UID)
+        self.assertEqual(f_info.st_gid, FAKE_GROUP_GID)
+
+
 @patch("charms.hpc_libs.v0.slurm_ops.subprocess.run")
 class TestCgroupConfig(TestCase):
     """Test the Slurmctld service cgroup config manager."""