feat(host_agent): a temperature sensor is judged by how it moves against the CPU (#5783)

A board's spare thermistor inputs report whatever an open input reads, and
the value rule from #5759 only catches the impossible ones. Once an hour,
sensors.refresh judges each sensor against the host's CPU (Package id N,
Tctl/Tdie/TccdN, PECI Agent N) over a week of hourly averages. A sensor is
ruled out if it falls as the CPU warms (r <= -0.5) or holds within 0.5 °C
while the CPU moved 10 °C or more. A judgment needs at least 72 shared hours.

The host temperature leaves judged sensors out at ingest. The host page
lists them under "Not counted", with each reason given once.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

tests/plugins/host_agent/test_expand_metrics.py

@@ -162,7 +162,8 @@ def test_the_page_lists_unwired_sensors_apart_hottest_first():
     """VDNT-Util01's sensor list read SYSTIN 115 °C in red at the top, though
     the host's temperature already left it out (#5759)."""
     from plugins.host_agent.routes import split_temps
-    real, unwired = split_temps({"SYSTIN": 115.0, "AUXTIN0": 110.0, "Package id 0": 52.0,
-                                 "Composite": 48.9, "PCH_CHIP_TEMP": 0.0, "CPUTIN": 47.0, "gone": None})
+    real, unwired = split_temps("h", {"SYSTIN": 115.0, "AUXTIN0": 110.0, "Package id 0": 52.0,
+                                      "Composite": 48.9, "PCH_CHIP_TEMP": 0.0, "CPUTIN": 47.0, "gone": None})
     assert list(real) == ["Package id 0", "Composite", "CPUTIN", "gone"]
     assert list(unwired) == ["SYSTIN", "AUXTIN0", "PCH_CHIP_TEMP"]
+    assert unwired["SYSTIN"] == (115.0, "reads a value no wired sensor gives")

tests/plugins/host_agent/test_host_health.py

@@ -77,3 +77,21 @@ async def test_the_metrics_page_grades_by_the_same_cutoffs(app):
     assert 'class="stat-val tone-warn" style="font-size:1.6rem;">72°C' in page
     assert '<span class="tone-crit" style="font-variant-numeric:tabular-nums;">90°C' in page
     assert not hex_colours(page), hex_colours(page)
+
+
+async def test_sensors_not_counted_are_listed_once_per_reason(app):
+    """Util01's five open inputs read as one list with its reason, not five
+    repeats of it; a sensor judged against the CPU carries its own (#5783)."""
+    held = "reads a value no wired sensor gives"
+    falls = "falls as the CPU warms (r -0.81 over 168 h)"
+    async with app.test_request_context("/plugins/host_agent/h1/metrics"):
+        page = await render_plugin_template(
+            app, "host_agent", "_host_metrics.html", host=SimpleNamespace(address="10.0.0.2"), stale=False,
+            reg=SimpleNamespace(distro=None, kernel=None, arch=None, agent_version=None,
+                                last_seen_at=datetime.now(timezone.utc)),
+            hostlvl={"temp_c_max": 52.0}, cores=[], nets={}, disks_io={}, temps={"Package id 0": 52.0},
+            unwired_temps={"SYSTIN": (115.0, held), "AUXTIN0": (110.0, held), "temp3": (62.0, falls)},
+            mounts={})
+    page = " ".join(page.split())
+    assert (f"Not counted, likely nothing wired to them: temp3 62°C ({falls}); "
+            f"SYSTIN 115°C, AUXTIN0 110°C ({held}).") in page

tests/plugins/host_agent/test_sensors.py

@@ -0,0 +1,74 @@
+"""Which temperature sensors are thermometers, judged from how each moves
+against the host's CPU (#5783). VDNT-Util01's SYSTIN ran at r = −0.81
+against its CPU package over a week; CPUTIN beside it at +0.83 (#5782)."""
+
+import math
+from datetime import datetime, timedelta, timezone
+
+from plugins.host_agent import sensors as s
+
+T0 = datetime(2026, 10, 8, tzinfo=timezone.utc)
+
+
+def _hours(n, f):
+    """{hour: f(i)} for n hourly buckets."""
+    return {T0 + timedelta(hours=i): f(i) for i in range(n)}
+
+
+def cpu(i):                                  # a CPU that swings 50-80 °C through the day
+    return 65 + 15 * math.sin(i / 24 * 2 * math.pi)
+
+
+def test_a_sensor_that_falls_as_the_cpu_warms_is_not_a_thermometer():
+    ref = _hours(96, cpu)
+    systin = _hours(96, lambda i: 114 - 0.1 * (cpu(i) - 65))       # Util01's shape
+    why = s.judge(ref, systin)
+    assert why.startswith("falls as the CPU warms (r -1.00 over 96 h")
+
+
+def test_a_sensor_that_holds_still_is_not_a_thermometer():
+    why = s.judge(_hours(96, cpu), _hours(96, lambda i: 110.0 + (0.1 if i % 2 else 0.0)))
+    assert why == "holds at 110 °C while the CPU moved 50–80 °C"
+
+
+def test_a_wired_sensor_is_left_alone_even_when_it_barely_follows_the_cpu():
+    ref = _hours(96, cpu)
+    assert s.judge(ref, _hours(96, lambda i: 40 + 0.3 * (cpu(i) - 65))) is None     # a board sensor
+    assert s.judge(ref, _hours(96, lambda i: 45 + 3 * math.sin(i))) is None          # a drive, its own rhythm
+
+
+def test_nothing_is_judged_without_enough_history_or_a_cpu_that_moved():
+    falls = lambda i: 114 - 0.1 * (cpu(i) - 65)                                      # noqa: E731
+    assert s.judge(_hours(48, cpu), _hours(48, falls)) is None                      # 48 h < MIN_HOURS
+    idle = _hours(96, lambda i: 50 + (i % 3))                                       # moved 2 °C
+    assert s.judge(idle, _hours(96, lambda i: 110 - (i % 3))) is None
+
+
+def test_the_cpu_reference_is_found_by_its_driver_names():
+    assert s.reference_label(["SYSTIN", "Core 0", "Package id 1", "Package id 0"]) == "Package id 0"
+    assert s.reference_label(["Tccd1", "Tctl", "Composite"]) == "Tctl"
+    assert s.reference_label(["PECI Agent 0", "SYSTIN"]) == "PECI Agent 0"
+    assert s.reference_label(["Composite", "temp1"]) is None
+
+
+def test_hosts_are_judged_from_hourly_rows_and_a_host_without_a_cpu_reading_is_skipped():
+    rows = []
+    for i in range(96):
+        h = T0 + timedelta(hours=i)
+        rows += [("util:temp:Package id 0", h, cpu(i)),
+                 ("util:temp:SYSTIN", h, 114 - 0.1 * (cpu(i) - 65)),
+                 ("util:temp:CPUTIN", h, 40 + 0.3 * (cpu(i) - 65)),
+                 ("nas:temp:temp1", h, 110 - 0.1 * (cpu(i) - 65)),
+                 ("util", h, 1.0)]                                                   # not a sensor row
+    got = s.judge_hosts(rows)
+    assert list(got) == ["util"] and list(got["util"]) == ["SYSTIN"]
+
+
+def test_a_judged_sensor_is_left_out_of_the_host_temperature_and_listed_with_why(monkeypatch):
+    from plugins.host_agent.routes import is_thermometer, split_temps
+    monkeypatch.setattr(s, "_unwired", {"util": {"temp3": "falls as the CPU warms (r -0.81 over 168 h)"}})
+    assert not is_thermometer("util", "temp3", 62.0)       # a plausible value, still not a thermometer
+    assert is_thermometer("other", "temp3", 62.0)
+    real, unwired = split_temps("util", {"temp3": 62.0, "Package id 0": 55.0})
+    assert list(real) == ["Package id 0"]
+    assert unwired == {"temp3": (62.0, "falls as the CPU warms (r -0.81 over 168 h)")}