Skip to content

use mig_strategy and numeric mig_instance_id for OCP MIG GPU rows #613

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
djnakabaale merged 4 commits into from
Mar 17, 2026
Merged

use mig_strategy and numeric mig_instance_id for OCP MIG GPU rows #613

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
4d2cfaf
use mig_strategy and numeric mig_instance_id for OCP MIG GPU rows, an…
djnakabaale Mar 17, 2026
e2fe97b
undo GPU_VENDOR change
djnakabaale Mar 17, 2026
3446c0c
apply error handling suggestion
djnakabaale Mar 17, 2026
bc3ab79
bump coverage
djnakabaale Mar 17, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion nise/__init__.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,5 +1,5 @@
from .helpers import gcp_calculate_persistent_disk_usage_amount
from .helpers import gcp_calculate_usage_amount_in_pricing

__version__ = "5.3.8"
__version__ = "5.3.9"
VERSION = __version__.split(".")
29 changes: 19 additions & 10 deletions nise/generators/ocp/ocp_generator.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -221,7 +221,7 @@
"gpu_pod_uptime",
"mig_instance_id",
"mig_profile",
"mig_slice_count",
"mig_strategy",
)
COST_OCP_REPORT_TYPE_TO_COLS = {
OCP_POD_USAGE: OCP_POD_USAGE_COLUMNS,
Expand Down Expand Up @@ -1655,27 +1655,36 @@ def _gen_gpus(self): # noqa: C901
"gpu_memory_capacity_mib": gpu_memory,
"mig_instance_id": None,
"mig_profile": None,
"mig_slice_count": None,
"mig_strategy": None,
}
)
continue

for mig_idx, mig_spec in enumerate(mig_instances):
mig_profile = mig_spec.get("mig_profile")
mig_slice_count = mig_spec.get("mig_slice_count")
if not all([mig_profile, mig_slice_count]):
raise ValueError(f"MIG instance for pod '{pod_name}' requires mig_profile and mig_slice_count")
mig_strategy = mig_spec.get("mig_strategy")
if not all([mig_profile, mig_strategy]):
raise ValueError(f"MIG instance for pod '{pod_name}' requires mig_profile and mig_strategy")
mig_name = f"nise.ocp.mig.{node_name}.{pod_name}.{gpu_idx}.{mig_idx}"
mig_instance_id = f"MIG-{uuid5(NAMESPACE_DNS, mig_name)}"
mig_uuid = f"MIG-{uuid5(NAMESPACE_DNS, mig_name)}"
raw_mig_instance_id = mig_spec.get("mig_instance_id", mig_idx + 1)
try:
# MIG instance IDs are numeric; default to 1-based per-GPU-slice index when not provided.
mig_instance_id = int(raw_mig_instance_id)
except (ValueError, TypeError) as exc:
raise ValueError(
f"pod {pod_name}: mig_instance_id must be an integer value, got {raw_mig_instance_id}"
) from exc

pod_gpus.append(
{
"gpu_uuid": mig_instance_id,
"gpu_uuid": mig_uuid,
"gpu_model_name": gpu_model,
"gpu_vendor_name": GPU_VENDOR,
"gpu_memory_capacity_mib": gpu_memory,
"mig_instance_id": mig_instance_id,
"mig_profile": mig_profile,
"mig_slice_count": mig_slice_count,
"mig_strategy": mig_strategy,
}
)

Expand Down Expand Up @@ -1718,7 +1727,7 @@ def _gen_hourly_gpu_usage(self, **kwargs):
gpu_pod_uptime=gpu_pod_uptime,
mig_instance_id=gpu.get("mig_instance_id"),
mig_profile=gpu.get("mig_profile"),
mig_slice_count=gpu.get("mig_slice_count"),
mig_strategy=gpu.get("mig_strategy"),
**kwargs,
)
yield row
Expand All @@ -1736,7 +1745,7 @@ def _update_gpu_data(self, row, start, end, **kwargs):
"gpu_pod_uptime": kwargs.get("gpu_pod_uptime"),
"mig_instance_id": kwargs.get("mig_instance_id"),
"mig_profile": kwargs.get("mig_profile"),
"mig_slice_count": kwargs.get("mig_slice_count"),
"mig_strategy": kwargs.get("mig_strategy"),
}
row.update(data)
return row
Expand Down
17 changes: 12 additions & 5 deletions tests/ocp_gpu_static_report.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -64,26 +64,33 @@ generators:
mig_instances:
- mig_instance:
mig_profile: "3g.40gb"
mig_slice_count: 3
mig_strategy: "mixed"
- mig_instance:
mig_profile: "3g.40gb"
mig_slice_count: 3
mig_strategy: "mixed"
- gpu:
gpu_model: "A100"
gpu_memory_capacity_mib: 40960
mig_instances:
- mig_instance:
mig_profile: "1g.5gb"
mig_slice_count: 1
mig_strategy: "mixed"
- mig_instance:
mig_profile: "2g.10gb"
mig_slice_count: 2
mig_strategy: "mixed"
- mig_instance:
mig_profile: "4g.20gb"
mig_slice_count: 4
mig_strategy: "mixed"
- gpu:
gpu_model: "H100"
gpu_memory_capacity_mib: 81920
- gpu:
gpu_model: "H100"
gpu_memory_capacity_mib: 81920
- gpu:
gpu_model: "GH100"
gpu_memory_capacity_mib: 40960
mig_instances:
- mig_instance:
mig_profile: "3g.40gb"
mig_strategy: "single"
40 changes: 25 additions & 15 deletions tests/test_ocp_generator.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1201,7 +1201,7 @@ def _mig_gpu_attributes(self, pod_name="mig-pod", gpu_overrides=None, mig_instan
"mig_instances": [
{
"mig_profile": "3g.40gb",
"mig_slice_count": 3,
"mig_strategy": "mixed",
},
],
}
Expand Down Expand Up @@ -1282,20 +1282,29 @@ def test_gen_gpus_yaml_with_mig_instances(self):
gpu = pod_gpus[0]
self.assertEqual(gpu["gpu_model_name"], "H100")
self.assertEqual(gpu["mig_profile"], "3g.40gb")
self.assertEqual(gpu["mig_slice_count"], 3)
self.assertIn("MIG-", gpu["mig_instance_id"])
self.assertEqual(gpu["gpu_uuid"], gpu["mig_instance_id"])
self.assertEqual(gpu["mig_strategy"], "mixed")
self.assertIn("MIG-", gpu["gpu_uuid"])
self.assertIsInstance(gpu["mig_instance_id"], int)
self.assertEqual(gpu["mig_instance_id"], 1)

def test_gen_gpus_raises_when_mig_instance_missing_required_fields(self):
"""Test that ValueError is raised when a MIG instance lacks mig_profile or mig_slice_count."""
"""Test that ValueError is raised when a MIG instance lacks mig_profile or mig_strategy."""
attrs = self._mig_gpu_attributes(pod_name="incomplete-mig-pod")
del attrs["nodes"][0]["namespaces"]["mig-namespace"]["pods"][0]["gpus"][0]["mig_instances"][0][
"mig_slice_count"
]
del attrs["nodes"][0]["namespaces"]["mig-namespace"]["pods"][0]["gpus"][0]["mig_instances"][0]["mig_strategy"]
with self.assertRaises(ValueError) as ctx:
OCPGenerator(self.two_hours_ago, self.now, attrs)
self.assertIn("mig_profile", str(ctx.exception))
self.assertIn("mig_slice_count", str(ctx.exception))
self.assertIn("mig_strategy", str(ctx.exception))

def test_gen_gpus_raises_when_mig_instance_id_is_not_integer(self):
"""Test that ValueError is raised when mig_instance_id cannot be coerced to int."""
attrs = self._mig_gpu_attributes(
pod_name="invalid-mig-id-pod", mig_instance_overrides={"mig_instance_id": "not-an-int"}
)
with self.assertRaises(ValueError) as ctx:
OCPGenerator(self.two_hours_ago, self.now, attrs)
self.assertIn("invalid-mig-id-pod", str(ctx.exception))
self.assertIn("mig_instance_id must be an integer value", str(ctx.exception))

def test_gen_hourly_gpu_usage_includes_mig_fields(self):
"""Test that GPU usage rows include MIG fields when pod has MIG instances."""
Expand All @@ -1306,8 +1315,9 @@ def test_gen_hourly_gpu_usage_includes_mig_fields(self):
self.assertGreater(len(gpu_data), 0)
row = gpu_data[0]
self.assertEqual(row["mig_profile"], "3g.40gb")
self.assertEqual(row["mig_slice_count"], 3)
self.assertIn("MIG-", row["mig_instance_id"])
self.assertEqual(row["mig_strategy"], "mixed")
self.assertIsInstance(row["mig_instance_id"], int)
self.assertEqual(row["mig_instance_id"], 1)

def test_gen_gpus_random_generation(self):
"""Test random GPU generation (10% of pods)."""
Expand Down Expand Up @@ -1387,9 +1397,9 @@ def test_update_gpu_data(self):
"gpu_vendor_name": GPU_VENDOR,
"gpu_memory_capacity_mib": 15360,
"gpu_pod_uptime": 3000.123456,
"mig_instance_id": "MIG-test-uuid",
"mig_instance_id": 1,
"mig_profile": "3g.40gb",
"mig_slice_count": 3,
"mig_strategy": "mixed",
}
updated_row = generator._update_gpu_data(row, self.two_hours_ago, self.now, **kwargs)
self.assertEqual(updated_row["node"], "test-node")
Expand All @@ -1400,9 +1410,9 @@ def test_update_gpu_data(self):
self.assertEqual(updated_row["gpu_vendor_name"], GPU_VENDOR)
self.assertEqual(updated_row["gpu_memory_capacity_mib"], 15360)
self.assertEqual(updated_row["gpu_pod_uptime"], 3000.123456)
self.assertEqual(updated_row["mig_instance_id"], "MIG-test-uuid")
self.assertEqual(updated_row["mig_instance_id"], 1)
self.assertEqual(updated_row["mig_profile"], "3g.40gb")
self.assertEqual(updated_row["mig_slice_count"], 3)
self.assertEqual(updated_row["mig_strategy"], "mixed")

def test_gpu_usage_with_multiple_gpus_per_pod(self):
"""Test that multiple GPUs per pod generate separate rows."""
Expand Down
Loading