fix: lock CPU power history swap

codecarbon/external/hardware.py

@@ -4,6 +4,7 @@
 
 import math
 import re
+import threading
 import time
 from abc import ABC, abstractmethod
 from dataclasses import dataclass
@@ -207,6 +208,8 @@ def __init__(
     ):
         assert tracking_mode in ["machine", "process"]
         self._power_history: List[Power] = []
+        # the monitor thread appends while the measurement thread drains
+        self._power_history_lock = threading.Lock()
         self._output_dir = output_dir
         self._mode = mode
         self._model = model
@@ -393,12 +396,12 @@ def _get_energy_from_cpus(self, delay: Time) -> Energy:
         return Energy.from_energy(energy)
 
     def total_power(self) -> Power:
-        self._power_history.append(self._get_power_from_cpus())
-        power_history_in_W = [power.W for power in self._power_history]
-        self._power_history = []
-        if not power_history_in_W:
-            logger.warning("No power samples collected, returning 0 W")
-            return Power.from_watts(0)
+        latest_sample = self._get_power_from_cpus()
+        with self._power_history_lock:
+            power_history = self._power_history
+            self._power_history = []
+        power_history.append(latest_sample)
+        power_history_in_W = [power.W for power in power_history]
         cpu_power = sum(power_history_in_W) / len(power_history_in_W)
         return Power.from_watts(cpu_power)
 
@@ -432,7 +435,8 @@ def start(self):
 
     def monitor_power(self):
         cpu_power = self._get_power_from_cpus()
-        self._power_history.append(cpu_power)
+        with self._power_history_lock:
+            self._power_history.append(cpu_power)
 
     def get_model(self):
         return self._model

tests/test_cpu_load.py

@@ -1,3 +1,4 @@
+import threading
 import unittest
 from time import sleep
 from unittest import mock
@@ -172,6 +173,40 @@ def test_cpu_total_power_fetches_sample_when_history_is_empty(
         mocked_get_power_from_cpus.assert_called_once()
         self.assertEqual(cpu._power_history, [])
 
+    @mock.patch(
+        "codecarbon.external.hardware.CPU._get_power_from_cpus",
+        return_value=Power.from_watts(1),
+    )
+    def test_cpu_total_power_keeps_samples_added_while_draining(
+        self,
+        mocked_get_power_from_cpus,
+        mocked_is_psutil_available,
+        mocked_is_powergadget_available,
+        mocked_is_rapl_available,
+    ):
+        """A sample appended by the monitor thread while total_power drains the
+        history must not be lost (see issue #1315)."""
+        cpu = CPU.from_utils(
+            None, MODE_CPU_LOAD, "Intel(R) Core(TM) i7-7600U CPU @ 2.80GHz", 100
+        )
+        cpu._power_history = [Power.from_watts(1)]
+        appended = threading.Event()
+
+        def monitor():
+            cpu.monitor_power()
+            appended.set()
+
+        monitor_thread = threading.Thread(target=monitor)
+        with cpu._power_history_lock:
+            monitor_thread.start()
+            # The monitor thread must wait instead of appending to a history
+            # total_power is about to discard.
+            self.assertFalse(appended.wait(0.2))
+        monitor_thread.join(1)
+
+        self.assertTrue(appended.is_set())
+        self.assertEqual(len(cpu._power_history), 2)
+
     @mock.patch(
         "codecarbon.external.hardware.CPU._get_power_from_cpus",
         return_value=Power.from_watts(30),