From c582bdd67337652b16de12390bcdc99679cde13c Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Heath=20Dutton=F0=9F=95=B4=EF=B8=8F?=
 <heathdutton@gmail.com>
Date: Sat, 20 Dec 2025 20:28:59 -0500
Subject: [PATCH] bugfix: Handle MacMon errors gracefully

---
 src/exo/worker/utils/macmon.py            | 18 ++++--
 src/exo/worker/utils/profile.py           |  1 +
 src/exo/worker/utils/tests/__init__.py    |  0
 src/exo/worker/utils/tests/test_macmon.py | 77 +++++++++++++++++++++++
 4 files changed, 90 insertions(+), 6 deletions(-)
 create mode 100644 src/exo/worker/utils/tests/__init__.py
 create mode 100644 src/exo/worker/utils/tests/test_macmon.py

diff --git a/src/exo/worker/utils/macmon.py b/src/exo/worker/utils/macmon.py
index 3e4e29e1..1e892d77 100644
--- a/src/exo/worker/utils/macmon.py
+++ b/src/exo/worker/utils/macmon.py
@@ -1,6 +1,7 @@
 import platform
 import shutil
 from subprocess import CalledProcessError
+from typing import cast
 
 from anyio import run_process
 from pydantic import BaseModel, ConfigDict, ValidationError
@@ -80,7 +81,6 @@ async def get_metrics_async() -> Metrics:
     """
     path = _get_binary_path()
 
-    result = None
     try:
         # TODO: Keep Macmon running in the background?
         result = await run_process([path, "pipe", "-s", "1"])
@@ -90,8 +90,14 @@ async def get_metrics_async() -> Metrics:
     except ValidationError as e:
         raise MacMonError(f"Error parsing JSON output: {e}") from e
     except CalledProcessError as e:
-        if result:
-            raise MacMonError(
-                f"MacMon failed with return code {result.returncode}"
-            ) from e
-        raise e
+        stderr_msg = "no stderr"
+        stderr_output = cast(bytes | str | None, e.stderr)
+        if stderr_output is not None:
+            stderr_msg = (
+                stderr_output.decode()
+                if isinstance(stderr_output, bytes)
+                else str(stderr_output)
+            )
+        raise MacMonError(
+            f"MacMon failed with return code {e.returncode}: {stderr_msg}"
+        ) from e
diff --git a/src/exo/worker/utils/profile.py b/src/exo/worker/utils/profile.py
index 30aca08c..316973ad 100644
--- a/src/exo/worker/utils/profile.py
+++ b/src/exo/worker/utils/profile.py
@@ -109,5 +109,6 @@ async def start_polling_node_metrics(
             )
         except MacMonError as e:
             logger.opt(exception=e).error("Resource Monitor encountered error")
+            return
         finally:
             await anyio.sleep(poll_interval_s)
diff --git a/src/exo/worker/utils/tests/__init__.py b/src/exo/worker/utils/tests/__init__.py
new file mode 100644
index 00000000..e69de29b
diff --git a/src/exo/worker/utils/tests/test_macmon.py b/src/exo/worker/utils/tests/test_macmon.py
new file mode 100644
index 00000000..afb19aa4
--- /dev/null
+++ b/src/exo/worker/utils/tests/test_macmon.py
@@ -0,0 +1,77 @@
+"""Tests for macmon error handling.
+
+These tests verify that MacMon errors are handled gracefully without
+crashing the application or spamming logs.
+"""
+
+import platform
+from subprocess import CalledProcessError
+from unittest.mock import AsyncMock, patch
+
+import pytest
+
+from exo.worker.utils.macmon import MacMonError, get_metrics_async
+
+
+@pytest.mark.skipif(
+    platform.system().lower() != "darwin" or "arm" not in platform.machine().lower(),
+    reason="MacMon only supports macOS with Apple Silicon",
+)
+class TestMacMonErrorHandling:
+    """Test MacMon error handling."""
+
+    async def test_called_process_error_wrapped_as_macmon_error(self) -> None:
+        """CalledProcessError should be wrapped as MacMonError."""
+        mock_error = CalledProcessError(
+            returncode=1,
+            cmd=["macmon", "pipe", "-s", "1"],
+            stderr=b"some error message",
+        )
+
+        with (
+            patch(
+                "exo.worker.utils.macmon.shutil.which", return_value="/usr/bin/macmon"
+            ),
+            patch(
+                "exo.worker.utils.macmon.run_process", new_callable=AsyncMock
+            ) as mock_run,
+        ):
+            mock_run.side_effect = mock_error
+
+            with pytest.raises(MacMonError) as exc_info:
+                await get_metrics_async()
+
+            assert "MacMon failed with return code 1" in str(exc_info.value)
+            assert "some error message" in str(exc_info.value)
+
+    async def test_called_process_error_with_no_stderr(self) -> None:
+        """CalledProcessError with no stderr should be handled gracefully."""
+        mock_error = CalledProcessError(
+            returncode=1,
+            cmd=["macmon", "pipe", "-s", "1"],
+            stderr=None,
+        )
+
+        with (
+            patch(
+                "exo.worker.utils.macmon.shutil.which", return_value="/usr/bin/macmon"
+            ),
+            patch(
+                "exo.worker.utils.macmon.run_process", new_callable=AsyncMock
+            ) as mock_run,
+        ):
+            mock_run.side_effect = mock_error
+
+            with pytest.raises(MacMonError) as exc_info:
+                await get_metrics_async()
+
+            assert "MacMon failed with return code 1" in str(exc_info.value)
+            assert "no stderr" in str(exc_info.value)
+
+    async def test_macmon_not_found_raises_macmon_error(self) -> None:
+        """When macmon is not found in PATH, MacMonError should be raised."""
+        with patch("exo.worker.utils.macmon.shutil.which", return_value=None):
+            with pytest.raises(MacMonError) as exc_info:
+                await get_metrics_async()
+
+            assert "MacMon not found in PATH" in str(exc_info.value)