fix(fitness): modernize types, fix async blocking, and improve error handling

- Replace typing.Optional/List/Dict with native annotations via __future__
- Convert subprocess.run to asyncio.create_subprocess_exec in _get_cuda_version
  to avoid blocking the event loop from async functions
- Add proper exception chaining (from e / from None) across all modules
- Remove broad except Exception catches that silently suppressed registration
  failures in _ensure_gpu_check_registered and _ensure_system_checks_registered
- Fix memory unit conversion bug in /proc/meminfo fallback path
- Add zero-division guard in _check_disk_space
- Re-raise RuntimeError in benchmark checks to propagate actual failures
- Update tests to mock asyncio.create_subprocess_exec instead of subprocess.run

Author: deanq

runpod/_binary_helpers.py

@@ -2,12 +2,13 @@
 Helper utilities for locating package-bundled binaries.
 """
 
+from __future__ import annotations
+
 import os
 from pathlib import Path
-from typing import Optional
 
 
-def get_binary_path(binary_name: str) -> Optional[Path]:
+def get_binary_path(binary_name: str) -> Path | None:
     """
     Locate a binary file within the runpod package.
 

runpod/serverless/modules/rp_fitness.py

@@ -8,18 +8,20 @@
 Fitness checks do NOT run in local development mode or testing mode.
 """
 
+from __future__ import annotations
+
 import inspect
 import sys
 import time
 import traceback
-from typing import Callable, List
+from collections.abc import Callable
 
 from .rp_logger import RunPodLogger
 
 log = RunPodLogger()
 
 # Global registry for fitness check functions, preserves registration order
-_fitness_checks: List[Callable] = []
+_fitness_checks: list[Callable] = []
 
 
 def register_fitness_check(func: Callable) -> Callable:
@@ -99,11 +101,7 @@ def _ensure_gpu_check_registered() -> None:
 
         auto_register_gpu_check()
     except ImportError:
-        # GPU fitness module not available
         log.debug("GPU fitness check module not found, skipping auto-registration")
-    except Exception as e:
-        # Don't fail fitness checks if auto-registration has issues
-        log.warn(f"Failed to auto-register GPU fitness check: {e}")
 
 
 def _ensure_system_checks_registered() -> None:
@@ -122,7 +120,9 @@ def _ensure_system_checks_registered() -> None:
 
     # Allow disabling system checks for testing
     if os.environ.get("RUNPOD_SKIP_AUTO_SYSTEM_CHECKS", "").lower() == "true":
-        log.debug("System fitness checks disabled via environment (RUNPOD_SKIP_AUTO_SYSTEM_CHECKS)")
+        log.debug(
+            "System fitness checks disabled via environment (RUNPOD_SKIP_AUTO_SYSTEM_CHECKS)"
+        )
         _system_checks_registered = True
         return
 
@@ -133,11 +133,7 @@ def _ensure_system_checks_registered() -> None:
 
         auto_register_system_checks()
     except ImportError:
-        # System fitness module not available
         log.debug("System fitness check module not found, skipping auto-registration")
-    except Exception as e:
-        # Don't fail fitness checks if auto-registration has issues
-        log.warn(f"Failed to auto-register system fitness checks: {e}")
 
 
 async def run_fitness_checks() -> None:
@@ -206,8 +202,7 @@ async def run_fitness_checks() -> None:
             full_traceback = traceback.format_exc()
 
             log.error(
-                f"Fitness check failed: {check_name} | "
-                f"{error_type}: {error_message}"
+                f"Fitness check failed: {check_name} | {error_type}: {error_message}"
             )
             log.debug(f"Traceback:\n{full_traceback}")
 

runpod/serverless/modules/rp_gpu_fitness.py

@@ -8,11 +8,13 @@
 Auto-registers when GPUs are detected, skips silently on CPU-only workers.
 """
 
+from __future__ import annotations
+
 import asyncio
 import os
 import subprocess
 from pathlib import Path
-from typing import Any, Dict, Optional
+from typing import Any
 
 from runpod._binary_helpers import get_binary_path
 from .rp_fitness import register_fitness_check
@@ -25,7 +27,7 @@
 MAX_ERROR_MESSAGES = int(os.environ.get("RUNPOD_GPU_MAX_ERROR_MESSAGES", "10"))
 
 
-def _get_gpu_test_binary_path() -> Optional[Path]:
+def _get_gpu_test_binary_path() -> Path | None:
     """
     Locate gpu_test binary in package.
 
@@ -35,7 +37,7 @@ def _get_gpu_test_binary_path() -> Optional[Path]:
     return get_binary_path("gpu_test")
 
 
-def _parse_gpu_test_output(output: str) -> Dict[str, Any]:
+def _parse_gpu_test_output(output: str) -> dict[str, Any]:
     """
     Parse gpu_test binary output and detect success/failure.
 
@@ -92,9 +94,7 @@ def _parse_gpu_test_output(output: str) -> Dict[str, Any]:
             passed_count += 1
 
         # Check for errors
-        if any(
-            err in line.lower() for err in ["failed", "error", "cannot", "unable"]
-        ):
+        if any(err in line.lower() for err in ["failed", "error", "cannot", "unable"]):
             result["errors"].append(line)
 
     result["gpu_count"] = passed_count
@@ -105,7 +105,7 @@ def _parse_gpu_test_output(output: str) -> Dict[str, Any]:
     return result
 
 
-async def _run_gpu_test_binary() -> Dict[str, Any]:
+async def _run_gpu_test_binary() -> dict[str, Any]:
     """
     Execute gpu_test binary and parse output.
 
@@ -211,11 +211,15 @@ def _run_gpu_test_fallback() -> None:
         )
 
     except FileNotFoundError:
-        raise RuntimeError("nvidia-smi not found. Cannot validate GPU availability.") from None
+        raise RuntimeError(
+            "nvidia-smi not found. Cannot validate GPU availability."
+        ) from None
     except subprocess.TimeoutExpired:
         raise RuntimeError("nvidia-smi timed out") from None
     except RuntimeError:
         raise
+    except Exception as e:
+        raise RuntimeError(f"nvidia-smi fallback check failed: {e}") from e
 
 
 async def _check_gpu_health() -> None: