From 514502e045071e080da95ebf205dac641b246352 Mon Sep 17 00:00:00 2001
From: adam <me@adamksmith.xyz>
Date: Tue, 16 Jun 2026 19:05:10 +0000
Subject: [PATCH] host_sensors: filter IPMI margin sensors; coretemp per-core
 CPU fallback

- skip IPMI '*Margin' sensors (thermal margin / distance-to-throttle reads in
  degrees C but is not an absolute temperature, e.g. robin's P1 Therm Margin)
- coretemp: when a CPU has no 'Package id' sensor (older Xeons like robin's
  X3430), fall back to the hottest Core as the CPU temp
---
 services/host_sensors.py | 28 +++++++++++++++++++---------
 1 file changed, 19 insertions(+), 9 deletions(-)

diff --git a/services/host_sensors.py b/services/host_sensors.py
index 9bdc191..4736f57 100644
--- a/services/host_sensors.py
+++ b/services/host_sensors.py
@@ -45,6 +45,10 @@ def _parse_ipmi(text: str) -> list[dict]:
         if len(parts) < 5:
             continue
         name, _sid, status, entity, reading = parts[:5]
+        # Skip relative/derived sensors (e.g. "P1 Therm Margin") — they read in
+        # "degrees C" but are a distance-to-throttle, not an absolute temp.
+        if "margin" in name.lower():
+            continue
         m = re.search(r"(-?\d+(?:\.\d+)?)\s*degrees?\s*C", reading, re.IGNORECASE)
         if not m:  # "no reading", "disabled", "ns", etc.
             continue
@@ -74,22 +78,28 @@ def read_coretemp() -> list[dict]:
                 continue
         except OSError:
             continue
+        packages: list[dict] = []
+        cores: list[int] = []
         for label_f in sorted(hw.glob("temp*_label")):
             try:
                 label = label_f.read_text().strip()
             except OSError:
                 continue
-            if not label.lower().startswith("package id"):
-                continue
             input_f = label_f.with_name(label_f.name.replace("_label", "_input"))
             try:
-                milli = int(input_f.read_text().strip())
+                temp = round(int(input_f.read_text().strip()) / 1000)
             except (OSError, ValueError):
                 continue
-            pkg = label.split("id")[-1].strip()
-            cpus.append({
-                "name": f"CPU {pkg}",
-                "temperature_c": round(milli / 1000),
-                "kind": "cpu",
-            })
+            low = label.lower()
+            if low.startswith("package id"):
+                pkg = label.split("id")[-1].strip()
+                packages.append({"name": f"CPU {pkg}", "temperature_c": temp, "kind": "cpu"})
+            elif low.startswith("core"):
+                cores.append(temp)
+        # Prefer the per-package sensor; older CPUs (no package sensor) fall back
+        # to the hottest core as the CPU temp.
+        if packages:
+            cpus.extend(packages)
+        elif cores:
+            cpus.append({"name": "CPU", "temperature_c": max(cores), "kind": "cpu"})
     return cpus