Record node status snapshots before runs

docs/guides/add-app.md

@@ -243,6 +243,19 @@ CI の build job では `scripts/build_tool_wrappers/` を `PATH` の先頭に
 
 この actual build snapshot は、将来の build cache key や、同じ source から異なる binary が生じた場合の原因確認に使う前提の記録です。
 
+### run placement / node status snapshot の方針
+
+CI の共通 job は、benchmark 本体の `run_start` より前に `results/node_status_snapshot_run.json` を記録します。
+app の `run.sh` からこの snapshot 用 helper を呼ぶ必要はありません。
+
+この snapshot は scheduler-neutral な診断情報で、scheduler kind、scheduler が示す node list、観測できた host 数、CPU 数、memory、load average、GPU の軽量状態、run 前に見える GPU compute process の件数と memory 合計を記録します。
+process ID、process name、user name は記録しません。
+SLURM では可能なら allocation 内の各 node で軽い worker を動かします。
+PBS / PJM / unknown scheduler では、取れる範囲の nodefile / environment / local host 情報だけを partial または unsupported として残します。
+
+Result JSON には hash、summary、Measurement Artifact への参照だけを入れ、詳細な node list や GPU 状態は `results/node_status_snapshot_run.json` 側に残します。
+Portal の public surface ではこの Measurement Artifact は表示しません。
+
 ### build cache の方針
 
 cross build job と native `build_run` job の build phase では、共通 wrapper `scripts/build_with_cache.sh` が `build.sh` の前後で build artifact cache を扱います。

result_server/routes/results_detail_routes.py

@@ -383,6 +383,17 @@ def _list_result_measurement_artifact_filenames(result, artifact_dir, *, include
         seen.add(filename)
         if os.path.isfile(os.path.join(artifact_dir, filename)):
             filenames.append(filename)
+    for artifact_path in _iter_result_node_status_artifact_paths(result):
+        filename = stored_measurement_artifact_filename_from_path(
+            timestamp,
+            result_uuid,
+            artifact_path,
+        )
+        if not filename or filename in seen:
+            continue
+        seen.add(filename)
+        if os.path.isfile(os.path.join(artifact_dir, filename)):
+            filenames.append(filename)
     return filenames
 
 
@@ -417,5 +428,17 @@ def _iter_result_timing_artifact_paths(result):
             yield path
 
 
+def _iter_result_node_status_artifact_paths(result):
+    node_status_snapshot = result.get("node_status_snapshot")
+    if not isinstance(node_status_snapshot, dict):
+        return
+    artifact = node_status_snapshot.get("artifact")
+    if not isinstance(artifact, dict) or artifact.get("type") != "file_reference":
+        return
+    path = _clean_result_value(artifact.get("path"))
+    if path:
+        yield path
+
+
 def _clean_result_value(value):
     return str(value or "").strip()

result_server/templates/result_detail.html

@@ -73,6 +73,10 @@
 {{ render_titled_key_value_table("PA Data Summary", profile_rows, "meta-table") }}
 {% endif %}
 
+{% if node_status_rows %}
+{{ render_titled_key_value_table("Node Status Snapshot", node_status_rows, "meta-table") }}
+{% endif %}
+
 {% if measurement_artifact_rows %}
 <div class="section">
     <h2>Measurement Artifacts</h2>

result_server/tests/test_measurement_artifacts.py

@@ -62,6 +62,11 @@ def test_builds_canonical_stored_names_for_legacy_profiles_and_timing_json():
         UUID,
         "results/qws_timing_CASE0.json",
     ) == f"measurement_artifact_{TIMESTAMP}_{UUID}_qws_timing_CASE0.json"
+    assert stored_measurement_artifact_filename_from_path(
+        TIMESTAMP,
+        UUID,
+        "results/node_status_snapshot_run.json",
+    ) == f"measurement_artifact_{TIMESTAMP}_{UUID}_node_status_snapshot_run.json"
 
 
 def test_profile_archive_candidates_keep_legacy_and_generic_names():
@@ -81,6 +86,7 @@ def test_profile_archive_candidates_keep_legacy_and_generic_names():
         f"padata_{TIMESTAMP}_{UUID}.tgz",
         f"padata_{TIMESTAMP}_{UUID}_padata_pairlist.tgz",
         f"measurement_artifact_{TIMESTAMP}_{UUID}_qws_timing_CASE0.json",
+        f"measurement_artifact_{TIMESTAMP}_{UUID}_node_status_snapshot_run.json",
     ],
 )
 def test_recognizes_served_measurement_artifact_filenames(filename):

result_server/tests/test_result_detail_template.py

@@ -141,6 +141,31 @@ def app():
             }
         ],
     },
+    "node_status_snapshot": {
+        "schema_version": 1,
+        "kind": "node_status_snapshot",
+        "collection_status": "ok",
+        "collection_warnings": [],
+        "scheduler_kind": "slurm",
+        "summary": {
+            "scheduler_host_count": 2,
+            "observed_host_count": 2,
+            "cpu_logical_counts": [64],
+            "memory_total_mib_min": 262144,
+            "memory_total_mib_max": 262144,
+            "memory_available_mib_min": 131072,
+            "load_average_1m_max": 0.5,
+            "load_average_5m_max": 0.8,
+            "observed_gpu_count": 4,
+            "gpu_memory_used_total_mib": 0,
+            "gpu_compute_process_count": 0,
+            "gpu_compute_memory_used_mib": 0,
+        },
+        "artifact": {
+            "type": "file_reference",
+            "path": "results/node_status_snapshot_run.json",
+        },
+    },
 }
 
 FULL_QUALITY = {
@@ -204,6 +229,20 @@ def test_meta_info_section(self, app):
         assert "build inputs hash matched" in html
         assert "rccs-cloud" in html
         assert "slurm" in html
+        assert "Node Status Snapshot" in html
+        assert "Hosts Observed" in html
+        assert "2/2" in html
+        assert "CPU Counts Observed" in html
+        assert "64" in html
+        assert "Host Memory Total" in html
+        assert "262144.000 MiB" in html
+        assert "Min Memory Available Before Run" in html
+        assert "131072.000 MiB" in html
+        assert "Max Load Average Before Run" in html
+        assert "1m=0.500; 5m=0.800" in html
+        assert "Compute Processes Before Run" in html
+        assert "0 process(es); 0.000 MiB" in html
+        assert "results/node_status_snapshot_run.json" in html
         assert "Timing Observations" in html
         assert "qws-case0-timers" in html
         assert "producer=qws" in html
@@ -233,6 +272,8 @@ def test_public_surface_meta_omits_operator_fields(self, app):
         assert "Cached Binary Created At" not in html
         assert "Allocation Project ID" not in html
         assert "Runner" not in html
+        assert "Node Status Snapshot" not in html
+        assert "node_status_snapshot_run" not in html
         assert "Timing Observations" not in html
         assert "qws_timing_CASE0" not in html
         assert "rccs-cloud" not in html
@@ -398,6 +439,48 @@ def test_timing_observation_artifact_is_hidden_on_public_surface(self, app):
         assert "Timing observation" not in html
         assert filename not in html
 
+    def test_node_status_snapshot_artifact_is_linked_on_console_surface(self, app):
+        result = {
+            **FULL_RESULT,
+            "_server_uuid": "12345678-1234-1234-1234-123456789abc",
+            "_server_timestamp": "20260819_161329",
+        }
+        filename = (
+            "measurement_artifact_20260819_161329_"
+            "12345678-1234-1234-1234-123456789abc_node_status_snapshot_run.json"
+        )
+
+        with app.test_request_context():
+            html = _render_result_detail(result, FULL_QUALITY, [filename])
+
+        assert "Measurement Artifacts" in html
+        assert "Node status snapshot" in html
+        assert "Run placement" in html
+        assert "results/node_status_snapshot_run.json" in html
+        assert f'href="/results/{filename}"' in html
+
+    def test_node_status_snapshot_artifact_is_hidden_on_public_surface(self, app):
+        result = {
+            **FULL_RESULT,
+            "_server_uuid": "12345678-1234-1234-1234-123456789abc",
+            "_server_timestamp": "20260819_161329",
+        }
+        filename = (
+            "measurement_artifact_20260819_161329_"
+            "12345678-1234-1234-1234-123456789abc_node_status_snapshot_run.json"
+        )
+
+        with app.test_request_context():
+            html = _render_result_detail(
+                result,
+                FULL_QUALITY,
+                [filename],
+                public_surface=True,
+            )
+
+        assert "Node status snapshot" not in html
+        assert filename not in html
+
     def test_vector_data_table(self, app):
         with app.test_request_context():
             html = _render_result_detail(FULL_RESULT, FULL_QUALITY)