fixed tests

Author: sanderegg

services/autoscaling/tests/unit/test_modules_cluster_scaling_dynamic.py

@@ -978,7 +978,7 @@ async def _assert_wait_for_ec2_instances_terminated() -> None:
             _ScaleUpParams(
                 imposed_instance_type=None,
                 service_resources=Resources(
-                    cpus=4, ram=TypeAdapter(ByteSize).validate_python("128Gib")
+                    cpus=4, ram=TypeAdapter(ByteSize).validate_python("114Gib")
                 ),
                 num_services=1,
                 expected_instance_type="r5n.4xlarge",
@@ -990,7 +990,7 @@ async def _assert_wait_for_ec2_instances_terminated() -> None:
             _ScaleUpParams(
                 imposed_instance_type="t2.xlarge",
                 service_resources=Resources(
-                    cpus=4, ram=TypeAdapter(ByteSize).validate_python("4Gib")
+                    cpus=2.6, ram=TypeAdapter(ByteSize).validate_python("4Gib")
                 ),
                 num_services=1,
                 expected_instance_type="t2.xlarge",
@@ -1002,7 +1002,7 @@ async def _assert_wait_for_ec2_instances_terminated() -> None:
             _ScaleUpParams(
                 imposed_instance_type="r5n.8xlarge",
                 service_resources=Resources(
-                    cpus=4, ram=TypeAdapter(ByteSize).validate_python("128Gib")
+                    cpus=4, ram=TypeAdapter(ByteSize).validate_python("114Gib")
                 ),
                 num_services=1,
                 expected_instance_type="r5n.8xlarge",
@@ -1165,7 +1165,7 @@ async def test_cluster_scaling_up_and_down_against_aws(
                 ),
                 num_services=10,
                 expected_instance_type="r5n.4xlarge",  # 1 GPU, 16 CPUs, 128GiB
-                expected_num_instances=4,
+                expected_num_instances=5,
             ),
             id="sim4life-light",
         ),
@@ -1254,7 +1254,7 @@ async def test_cluster_scaling_up_starts_multiple_instances(
             _ScaleUpParams(
                 imposed_instance_type="g4dn.2xlarge",  # 1 GPU, 8 CPUs, 32GiB
                 service_resources=Resources(
-                    cpus=8, ram=TypeAdapter(ByteSize).validate_python("15Gib")
+                    cpus=6.6, ram=TypeAdapter(ByteSize).validate_python("15Gib")
                 ),
                 num_services=12,
                 expected_instance_type="g4dn.2xlarge",  # 1 GPU, 8 CPUs, 32GiB
@@ -1263,7 +1263,7 @@ async def test_cluster_scaling_up_starts_multiple_instances(
             _ScaleUpParams(
                 imposed_instance_type="g4dn.8xlarge",  # 32CPUs, 128GiB
                 service_resources=Resources(
-                    cpus=32, ram=TypeAdapter(ByteSize).validate_python("20480MB")
+                    cpus=30.6, ram=TypeAdapter(ByteSize).validate_python("20480MB")
                 ),
                 num_services=7,
                 expected_instance_type="g4dn.8xlarge",  # 32CPUs, 128GiB
@@ -1556,7 +1556,7 @@ async def test_cluster_adapts_machines_on_the_fly(  # noqa: PLR0915
             _ScaleUpParams(
                 imposed_instance_type=None,
                 service_resources=Resources(
-                    cpus=4, ram=TypeAdapter(ByteSize).validate_python("128Gib")
+                    cpus=4, ram=TypeAdapter(ByteSize).validate_python("114Gib")
                 ),
                 num_services=1,
                 expected_instance_type="r5n.4xlarge",