Easier and more self documenting patterns for loading/saving Persistent
[python_utils.git] / remote_worker.py
index b58c6ba0a66f8d32b2b81af72a66d23493c9b2e5..8bc254070c7ec030a967efb8938e42639eaa5231 100755 (executable)
@@ -1,5 +1,7 @@
 #!/usr/bin/env python3
 
+# © Copyright 2021-2022, Scott Gasch
+
 """A simple utility to unpickle some code, run it, and pickle the
 results.
 """
@@ -7,9 +9,10 @@ results.
 import logging
 import os
 import signal
-import threading
 import sys
+import threading
 import time
+from typing import Optional
 
 import cloudpickle  # type: ignore
 import psutil  # type: ignore
@@ -17,9 +20,9 @@ import psutil  # type: ignore
 import argparse_utils
 import bootstrap
 import config
+from stopwatch import Timer
 from thread_utils import background_thread
 
-
 logger = logging.getLogger(__file__)
 
 cfg = config.add_commandline_args(
@@ -58,14 +61,12 @@ def watch_for_cancel(terminate_event: threading.Event) -> None:
         for ancestor in ancestors:
             name = ancestor.name()
             pid = ancestor.pid
-            logger.debug(f'Ancestor process {name} (pid={pid})')
+            logger.debug('Ancestor process %s (pid=%d)', name, pid)
             if 'ssh' in name.lower():
                 saw_sshd = True
                 break
         if not saw_sshd:
-            logger.error(
-                'Did not see sshd in our ancestors list?!  Committing suicide.'
-            )
+            logger.error('Did not see sshd in our ancestors list?!  Committing suicide.')
             os.system('pstree')
             os.kill(os.getpid(), signal.SIGTERM)
             time.sleep(5.0)
@@ -76,62 +77,67 @@ def watch_for_cancel(terminate_event: threading.Event) -> None:
         time.sleep(1.0)
 
 
+def cleanup_and_exit(
+    thread: Optional[threading.Thread],
+    stop_thread: Optional[threading.Event],
+    exit_code: int,
+) -> None:
+    if stop_thread is not None:
+        stop_thread.set()
+        assert thread is not None
+        thread.join()
+    sys.exit(exit_code)
+
+
 @bootstrap.initialize
 def main() -> None:
     in_file = config.config['code_file']
     out_file = config.config['result_file']
 
+    thread = None
     stop_thread = None
     if config.config['watch_for_cancel']:
         (thread, stop_thread) = watch_for_cancel()
 
-    logger.debug(f'Reading {in_file}.')
+    logger.debug('Reading %s.', in_file)
     try:
         with open(in_file, 'rb') as rb:
             serialized = rb.read()
     except Exception as e:
         logger.exception(e)
-        logger.critical(f'Problem reading {in_file}.  Aborting.')
-        stop_thread.set()
-        sys.exit(-1)
+        logger.critical('Problem reading %s. Aborting.', in_file)
+        cleanup_and_exit(thread, stop_thread, 1)
 
-    logger.debug(f'Deserializing {in_file}.')
+    logger.debug('Deserializing %s', in_file)
     try:
         fun, args, kwargs = cloudpickle.loads(serialized)
     except Exception as e:
         logger.exception(e)
-        logger.critical(f'Problem deserializing {in_file}.  Aborting.')
-        stop_thread.set()
-        sys.exit(-1)
+        logger.critical('Problem deserializing %s. Aborting.', in_file)
+        cleanup_and_exit(thread, stop_thread, 2)
 
     logger.debug('Invoking user code...')
-    start = time.time()
-    ret = fun(*args, **kwargs)
-    end = time.time()
-    logger.debug(f'User code took {end - start:.1f}s')
+    with Timer() as t:
+        ret = fun(*args, **kwargs)
+    logger.debug('User code took %.1fs', t())
 
     logger.debug('Serializing results')
     try:
         serialized = cloudpickle.dumps(ret)
     except Exception as e:
         logger.exception(e)
-        logger.critical(f'Could not serialize result ({type(ret)}).  Aborting.')
-        stop_thread.set()
-        sys.exit(-1)
+        logger.critical('Could not serialize result (%s). Aborting.', type(ret))
+        cleanup_and_exit(thread, stop_thread, 3)
 
-    logger.debug(f'Writing {out_file}.')
+    logger.debug('Writing %s', out_file)
     try:
         with open(out_file, 'wb') as wb:
             wb.write(serialized)
     except Exception as e:
         logger.exception(e)
-        logger.critical(f'Error writing {out_file}.  Aborting.')
-        stop_thread.set()
-        sys.exit(-1)
-
-    if stop_thread is not None:
-        stop_thread.set()
-        thread.join()
+        logger.critical('Error writing %s. Aborting.', out_file)
+        cleanup_and_exit(thread, stop_thread, 4)
+    cleanup_and_exit(thread, stop_thread, 0)
 
 
 if __name__ == '__main__':