Hotfix/v0.4.1

.github/workflows/ci.yml

@@ -0,0 +1,26 @@
+name: CI
+
+on:
+  pull_request:
+  push:
+    branches:
+      - main
+
+jobs:
+  test:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout code
+        uses: actions/checkout@v4
+
+      - name: Install uv
+        uses: astral-sh/setup-uv@v5
+
+      - name: Set up Python
+        run: uv python install 3.10
+
+      - name: Install dependencies
+        run: uv sync --dev
+
+      - name: Run tests
+        run: PYTHONPATH=src uv run pytest -q

CHANGELOG.md

@@ -7,6 +7,16 @@ and this project follows [Semantic Versioning](https://semver.org/spec/v2.0.0.ht
 
 ## [Unreleased]
 
+## [0.4.1] - 2026-02-23
+
+### Added
+- Regression tests for GPU parsing in `slurm_backend`, including realistic `scontrol show node -o` fixture coverage and obfuscated `squeue` job cases.
+- CI workflow running tests on pull requests and pushes to `main`.
+
+### Fixed
+- Correct GPU parsing for typed GRES values such as `gpu:h100:1` and `gres/gpu:h100:1` so model digits are not misread as GPU counts.
+- Preserve per-node GPU semantics in job parsing (`per-node * nodes`) while keeping `gpu_total`/`total_gpu` as explicit totals.
+
 ## [0.4.0] - 2026-02-23
 
 ### Added
@@ -62,7 +72,8 @@ and this project follows [Semantic Versioning](https://semver.org/spec/v2.0.0.ht
 ### Added
 - Initial `smon` release.
 
-[Unreleased]: https://github.com/RuHae/smon/compare/v0.4.0...HEAD
+[Unreleased]: https://github.com/RuHae/smon/compare/v0.4.1...HEAD
+[0.4.1]: https://github.com/RuHae/smon/compare/v0.4.0...v0.4.1
 [0.4.0]: https://github.com/RuHae/smon/compare/v0.3.0...v0.4.0
 [0.3.0]: https://github.com/RuHae/smon/compare/v0.2.0...v0.3.0
 [0.2.0]: https://github.com/RuHae/smon/compare/v0.1.0...v0.2.0

pyproject.toml

@@ -18,6 +18,7 @@ dependencies = [
 [dependency-groups]
 dev = [
     "pyinstaller>=6.18.0",
+    "pytest>=8.4.0",
 ]
 
 [project.scripts]

src/slurm_backend.py

@@ -4,6 +4,36 @@
 from fake_slurm_fixtures import run_fake_slurm_command
 from smon_config import USE_FAKE_DATA
 
+
+GPU_GRES_COUNT_PATTERN = re.compile(
+    r"(?:^|,)\s*(?:gres/)?gpu:(?:(?:[^:,()\s]+):)?(\d+)(?=$|,|\()"
+)
+GPU_ALLOC_TRES_PATTERN = re.compile(r"(?:^|,)\s*gres/gpu(?::[^=,\s]+)?=(\d+)(?=$|,)")
+GPU_TOTAL_PATTERN = re.compile(r"(?:gpu_total|total_gpu)[:=](\d+)")
+
+
+def _parse_gres_gpu_count(gres_value: str) -> int:
+    return sum(int(count) for count in GPU_GRES_COUNT_PATTERN.findall(gres_value))
+
+
+def _parse_alloc_tres_gpu_count(alloc_tres: str) -> int:
+    return sum(int(count) for count in GPU_ALLOC_TRES_PATTERN.findall(alloc_tres))
+
+
+def _parse_gpu_total(gpu_field: str) -> int | None:
+    match = GPU_TOTAL_PATTERN.search(gpu_field)
+    if match:
+        return int(match.group(1))
+    return None
+
+
+def _parse_gpu_per_node(gpu_field: str) -> int | None:
+    counts = GPU_GRES_COUNT_PATTERN.findall(gpu_field)
+    if counts:
+        return sum(int(count) for count in counts)
+    return None
+
+
 def run_slurm_command(cmd: str) -> str:
     if USE_FAKE_DATA:
         return run_fake_slurm_command(cmd)
@@ -35,18 +65,10 @@ def get_cluster_stats():
         m_u = int(data.get("AllocMem", 0))
         m_t = int(data.get("RealMemory", 1))
 
-        g_t, g_u = 0, 0
         gres_str = data.get("Gres", "")
-        if "gpu" in gres_str:
-            parts = re.findall(r":(\d+)", gres_str)
-            if parts:
-                g_t = int(parts[0])
-
         alloc_tres = data.get("AllocTRES", "")
-        if "gres/gpu" in alloc_tres:
-            match = re.search(r"gres/gpu[^=]*=(\d+)", alloc_tres)
-            if match:
-                g_u = int(match.group(1))
+        g_t = _parse_gres_gpu_count(gres_str)
+        g_u = _parse_alloc_tres_gpu_count(alloc_tres)
 
         t_cpu_u += c_u
         t_cpu_t += c_t
@@ -97,22 +119,18 @@ def get_job_stats():
 
         gpu_count = "-"
         gpu_field = parts[8]
-        if "gpu" in gpu_field:
-            try:
-                # Support explicit total notation in fixtures: gpu_total=64
-                total_match = re.search(
-                    r"(?:gpu_total|total_gpu|gres/gpu)[:=](\d+)", gpu_field
-                )
-                if total_match:
-                    gpu_count = total_match.group(1)
-                else:
+        try:
+            # Fixture compatibility: explicit total markers are whole-job totals.
+            explicit_total = _parse_gpu_total(gpu_field)
+            if explicit_total is not None:
+                gpu_count = str(explicit_total)
+            else:
+                per_node = _parse_gpu_per_node(gpu_field)
+                if per_node is not None:
                     node_mult = int(parts[6])
-                    per_node_match = re.search(r"gpu[^0-9]*(\d+)", gpu_field)
-                    if per_node_match:
-                        per_node = int(per_node_match.group(1))
-                        gpu_count = str(node_mult * per_node)
-            except Exception:
-                pass
+                    gpu_count = str(node_mult * per_node)
+        except Exception:
+            pass
 
         dep = parts[16]
         if dep == "(null)" or dep == "N/A":

tests/conftest.py

@@ -0,0 +1,6 @@
+from pathlib import Path
+import sys
+
+SRC_PATH = Path(__file__).resolve().parents[1] / "src"
+if str(SRC_PATH) not in sys.path:
+    sys.path.insert(0, str(SRC_PATH))

tests/fixture_loader.py

@@ -0,0 +1,11 @@
+from __future__ import annotations
+
+import json
+from pathlib import Path
+
+FIXTURE_PATH = Path(__file__).resolve().parents[1] / "fixtures" / "scontrol.json"
+
+
+def load_scontrol_show_node_output() -> str:
+    data = json.loads(FIXTURE_PATH.read_text())
+    return data["show_node"]

tests/test_slurm_backend_job_parsing.py

@@ -0,0 +1,31 @@
+from __future__ import annotations
+
+import slurm_backend
+
+
+def _build_squeue_output() -> str:
+    header = (
+        "JOBID USER STATE TIME LEFT PRIO NODES REASON GRES NAME CPU MEM PART "
+        "ACCOUNT QOS SUBMIT DEP"
+    )
+    rows = [
+        "100001 user01 RUNNING 00:10:00 00:50:00 100 1 cn01 gres/gpu:h100:4 train_a 64 256G all acct01 normal 2026-02-23T10:00:00 (null)",
+        "100002 user02 RUNNING 00:15:00 00:45:00 110 2 cn02 gres/gpu:h100:1 train_b 64 256G all acct02 normal 2026-02-23T10:05:00 (null)",
+        "100003 user03 RUNNING 00:05:00 00:55:00 120 1 cn03 gres/gpu:1 train_c 64 256G all acct03 normal 2026-02-23T10:10:00 (null)",
+        "100004 user04 RUNNING 00:20:00 00:40:00 130 4 cn04 gpu_total=64 train_d 64 256G all acct04 normal 2026-02-23T10:15:00 (null)",
+        "100005 user05 RUNNING 00:25:00 00:35:00 140 1 cn05 (null) train_e 64 256G all acct05 normal 2026-02-23T10:20:00 (null)",
+    ]
+    return "\n".join([header, *rows])
+
+
+def test_get_job_stats_parses_typed_and_untyped_gpu_counts(monkeypatch):
+    monkeypatch.setattr(slurm_backend, "run_slurm_command", lambda _cmd: _build_squeue_output())
+
+    jobs = slurm_backend.get_job_stats()
+    jobs_by_id = {job["id"]: job for job in jobs}
+
+    assert jobs_by_id["100001"]["gpu"] == "4"
+    assert jobs_by_id["100002"]["gpu"] == "2"
+    assert jobs_by_id["100003"]["gpu"] == "1"
+    assert jobs_by_id["100004"]["gpu"] == "64"
+    assert jobs_by_id["100005"]["gpu"] == "-"

tests/test_slurm_backend_node_parsing.py

@@ -0,0 +1,39 @@
+from __future__ import annotations
+
+import slurm_backend
+
+from fixture_loader import load_scontrol_show_node_output
+
+
+def test_get_cluster_stats_parses_real_scontrol_fixture(monkeypatch):
+    show_node_output = load_scontrol_show_node_output()
+    monkeypatch.setattr(slurm_backend, "run_slurm_command", lambda _cmd: show_node_output)
+
+    nodes, theoretical, real = slurm_backend.get_cluster_stats()
+
+    assert len(nodes) == 35
+    assert theoretical == (2218, 3920, 230, 280)
+    assert real == (3920, 280)
+
+
+def test_get_cluster_stats_parses_representative_gpu_nodes(monkeypatch):
+    show_node_output = load_scontrol_show_node_output()
+    monkeypatch.setattr(slurm_backend, "run_slurm_command", lambda _cmd: show_node_output)
+
+    nodes, _theoretical, _real = slurm_backend.get_cluster_stats()
+    by_name = {node["name"]: node for node in nodes}
+
+    assert by_name["cn01"]["g_t"] == 8
+    assert by_name["cn01"]["g_u"] == 6
+
+    assert by_name["cn02"]["g_t"] == 8
+    assert by_name["cn02"]["g_u"] == 4
+
+    assert by_name["cn06"]["g_t"] == 8
+    assert by_name["cn06"]["g_u"] == 3
+
+    assert by_name["cn20"]["g_t"] == 8
+    assert by_name["cn20"]["g_u"] == 1
+
+    assert by_name["cn13"]["g_u"] == 0
+    assert by_name["cn18"]["g_u"] == 0