[Validators] Add validator to prevent the use of Placement Groups with Capacity Blocks.

Signed-off-by: Giacomo Marciani <mgiacomo@amazon.com>

Author: gmarciani

CHANGELOG.md

@@ -8,6 +8,7 @@ CHANGELOG
 - Add new configuration section `Scheduling/SlurmSettings/ExternalSlurmdbd` to connect the cluster to an external Slurmdbd.
 - Add support for Amazon Linux 2023.
 - Add support for `price-capacity-optimized` as an `AllocationStrategy`.
+- Add validator to prevent the use of Placement Groups with Capacity Blocks.
 
 **BUG FIXES**
 - Fix DRA configuration to make `AutoExportPolicy` and `AutoImportPolicy` optional.

cli/src/pcluster/config/cluster_config.py

@@ -146,6 +146,7 @@
     InstanceTypeValidator,
     KeyPairValidator,
     PlacementGroupCapacityReservationValidator,
+    PlacementGroupCapacityTypeValidator,
     PlacementGroupNamingValidator,
 )
 from pcluster.validators.efs_validators import EfsMountOptionsValidator
@@ -3027,6 +3028,11 @@ def _register_validators(self, context: ValidatorContext = None):  # noqa: C901
                     os=self.image.os,
                     architecture=self.head_node.architecture,
                 )
+                self._register_validator(
+                    PlacementGroupCapacityTypeValidator,
+                    capacity_type=queue.capacity_type,
+                    placement_group_enabled=queue.is_placement_group_enabled_for_compute_resource(compute_resource),
+                )
                 # The validation below has to be in cluster config class instead of queue class
                 # to make sure the subnet APIs are cached by previous validations.
                 cr_target = compute_resource.capacity_reservation_target or queue.capacity_reservation_target

cli/src/pcluster/validators/ec2_validators.py

@@ -585,6 +585,20 @@ def _validate_with_subnets(
             )
 
 
+class PlacementGroupCapacityTypeValidator(Validator):
+    """Validate the placement group is compatible with the capacity type."""
+
+    def _validate(self, capacity_type: str, placement_group_enabled: bool):
+        if capacity_type == CapacityType.CAPACITY_BLOCK and placement_group_enabled:
+            self._add_failure(
+                "When using a capacity block reservation, a placement group constraint should not be set "
+                "as insufficient capacity errors may occur due to placement constraints outside of the "
+                "reservation even if the capacity reservation has remaining capacity. "
+                "Please remove the placement group for the compute resource.",
+                FailureLevel.WARNING,
+            )
+
+
 class PlacementGroupCapacityReservationValidator(Validator):
     """Validate the placement group is compatible with the capacity reservation target."""
 

cli/tests/pcluster/config/test_cluster_config.py

@@ -37,6 +37,7 @@
     SlurmSettings,
     Tag,
 )
+from pcluster.validators.ec2_validators import PlacementGroupCapacityTypeValidator
 from tests.pcluster.aws.dummy_aws_api import mock_aws_api
 
 mock_compute_resources = [
@@ -381,7 +382,14 @@ def test_registration_of_validators(self, memory_scheduling_enabled, mocker):
             ),
         )
         cluster_config._register_validators()
-        assert_that(cluster_config._validators).is_not_empty()
+        actual_validators = cluster_config._validators
+        assert_that(actual_validators).is_not_empty()
+        assert_that(actual_validators).contains(
+            (
+                PlacementGroupCapacityTypeValidator().__class__,
+                {"capacity_type": CapacityType.ONDEMAND, "placement_group_enabled": False},
+            )
+        )
 
     def test_instances_in_slurm_queue(self):
         queue = SlurmQueue(

cli/tests/pcluster/validators/test_ec2_validators.py

@@ -22,6 +22,7 @@
 from pcluster.aws.common import AWSClientError
 from pcluster.config.cluster_config import CapacityReservationTarget, PlacementGroup
 from pcluster.config.common import CapacityType
+from pcluster.validators.common import FailureLevel
 from pcluster.validators.ec2_validators import (
     AmiOsCompatibleValidator,
     CapacityReservationResourceGroupValidator,
@@ -36,10 +37,11 @@
     InstanceTypeValidator,
     KeyPairValidator,
     PlacementGroupCapacityReservationValidator,
+    PlacementGroupCapacityTypeValidator,
     PlacementGroupNamingValidator,
 )
 from tests.pcluster.aws.dummy_aws_api import mock_aws_api
-from tests.pcluster.validators.utils import assert_failure_messages
+from tests.pcluster.validators.utils import assert_failure_level, assert_failure_messages
 
 
 @pytest.mark.parametrize(
@@ -1380,6 +1382,64 @@ def test_placement_group_capacity_reservation_validator(
     assert_failure_messages(actual_failure, expected_message)
 
 
+@pytest.mark.parametrize(
+    "capacity_type, placement_group_enabled, expected_failure_level, expected_message",
+    [
+        (
+            CapacityType.ONDEMAND,
+            True,
+            None,
+            None,
+        ),
+        (
+            CapacityType.ONDEMAND,
+            False,
+            None,
+            None,
+        ),
+        (
+            CapacityType.SPOT,
+            True,
+            None,
+            None,
+        ),
+        (
+            CapacityType.SPOT,
+            False,
+            None,
+            None,
+        ),
+        (
+            CapacityType.CAPACITY_BLOCK,
+            True,
+            FailureLevel.WARNING,
+            "When using a capacity block reservation, a placement group constraint should not be set "
+            "as insufficient capacity errors may occur due to placement constraints outside of the "
+            "reservation even if the capacity reservation has remaining capacity. "
+            "Please remove the placement group for the compute resource.",
+        ),
+        (
+            CapacityType.CAPACITY_BLOCK,
+            False,
+            None,
+            None,
+        ),
+    ],
+)
+def test_placement_group_capacity_type_validator(
+    capacity_type,
+    placement_group_enabled,
+    expected_failure_level,
+    expected_message,
+):
+    actual_failures = PlacementGroupCapacityTypeValidator().execute(
+        capacity_type=capacity_type,
+        placement_group_enabled=placement_group_enabled,
+    )
+    assert_failure_level(actual_failures, expected_failure_level)
+    assert_failure_messages(actual_failures, expected_message)
+
+
 @pytest.mark.parametrize(
     "instance_type, instance_type_data, expected_message, logger_message",
     [