flwr-nightly 1.15.0.dev20250115__tar.gz → 1.15.0.dev20250117__tar.gz

{flwr_nightly-1.15.0.dev20250115 → flwr_nightly-1.15.0.dev20250117}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: flwr-nightly
-Version: 1.15.0.dev20250115
+Version: 1.15.0.dev20250117
 Summary: Flower: A Friendly Federated AI Framework
 Home-page: https://flower.ai
 License: Apache-2.0
@@ -32,8 +32,8 @@ Classifier: Topic :: Software Development :: Libraries :: Python Modules
 Classifier: Typing :: Typed
 Provides-Extra: rest
 Provides-Extra: simulation
-Requires-Dist: cryptography (>=42.0.4,<43.0.0)
-Requires-Dist: grpcio (>=1.60.0,<2.0.0,!=1.64.2,<=1.64.3)
+Requires-Dist: cryptography (>=43.0.1,<44.0.0)
+Requires-Dist: grpcio (>=1.69.0,<2.0.0)
 Requires-Dist: iterators (>=0.0.2,<0.0.3)
 Requires-Dist: numpy (>=1.26.0,<3.0.0)
 Requires-Dist: pathspec (>=0.12.1,<0.13.0)

{flwr_nightly-1.15.0.dev20250115 → flwr_nightly-1.15.0.dev20250117}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "poetry.core.masonry.api"
 
 [tool.poetry]
 name = "flwr-nightly"
-version = "1.15.0.dev20250115"
+version = "1.15.0.dev20250117"
 description = "Flower: A Friendly Federated AI Framework"
 license = "Apache-2.0"
 authors = ["The Flower Authors <hello@flower.ai>"]
@@ -68,9 +68,9 @@ flower-client-app = "flwr.client.supernode:run_client_app" # Deprecated
 python = "^3.9"
 # Mandatory dependencies
 numpy = ">=1.26.0,<3.0.0"
-grpcio = "^1.60.0,!=1.64.2,<=1.64.3"
+grpcio = "^1.69.0"
 protobuf = "^4.25.2"
-cryptography = "^42.0.4"
+cryptography = "^43.0.1"
 pycryptodome = "^3.18.0"
 iterators = "^0.0.2"
 typer = "^0.12.5"

{flwr_nightly-1.15.0.dev20250115 → flwr_nightly-1.15.0.dev20250117}/src/py/flwr/client/app.py

@@ -16,11 +16,11 @@
 
 
 import multiprocessing
-import signal
+import os
 import sys
+import threading
 import time
 from contextlib import AbstractContextManager
-from dataclasses import dataclass
 from logging import ERROR, INFO, WARN
 from os import urandom
 from pathlib import Path
@@ -346,10 +346,7 @@ def start_client_internal(
         transport, server_address
     )
 
-    app_state_tracker = _AppStateTracker()
-
     def _on_sucess(retry_state: RetryState) -> None:
-        app_state_tracker.is_connected = True
         if retry_state.tries > 1:
             log(
                 INFO,
@@ -359,7 +356,6 @@ def start_client_internal(
             )
 
     def _on_backoff(retry_state: RetryState) -> None:
-        app_state_tracker.is_connected = False
         if retry_state.tries == 1:
             log(WARN, "Connection attempt failed, retrying...")
         else:
@@ -396,7 +392,7 @@ def start_client_internal(
 
     runs: dict[int, Run] = {}
 
-    while not app_state_tracker.interrupt:
+    while True:
         sleep_duration: int = 0
         with connection(
             address,
@@ -435,9 +431,8 @@ def start_client_internal(
                         node_config=node_config,
                     )
 
-            app_state_tracker.register_signal_handler()
             # pylint: disable=too-many-nested-blocks
-            while not app_state_tracker.interrupt:
+            while True:
                 try:
                     # Receive
                     message = receive()
@@ -553,7 +548,7 @@ def start_client_internal(
 
                                 proc = mp_spawn_context.Process(
                                     target=_run_flwr_clientapp,
-                                    args=(command,),
+                                    args=(command, os.getpid()),
                                     daemon=True,
                                 )
                                 proc.start()
@@ -595,10 +590,7 @@ def start_client_internal(
                             e_code = ErrorCode.LOAD_CLIENT_APP_EXCEPTION
                             exc_entity = "SuperNode"
 
-                        if not app_state_tracker.interrupt:
-                            log(
-                                ERROR, "%s raised an exception", exc_entity, exc_info=ex
-                            )
+                        log(ERROR, "%s raised an exception", exc_entity, exc_info=ex)
 
                         # Create error message
                         reply_message = message.create_error_reply(
@@ -624,19 +616,14 @@ def start_client_internal(
                         run_id,
                     )
                     log(INFO, "")
-
-                except StopIteration:
-                    sleep_duration = 0
-                    break
             # pylint: enable=too-many-nested-blocks
 
             # Unregister node
-            if delete_node is not None and app_state_tracker.is_connected:
+            if delete_node is not None:
                 delete_node()  # pylint: disable=not-callable
 
         if sleep_duration == 0:
             log(INFO, "Disconnect and shut down")
-            del app_state_tracker
             break
 
         # Sleep and reconnect afterwards
@@ -812,24 +799,17 @@ def _init_connection(transport: Optional[str], server_address: str) -> tuple[
     return connection, address, error_type
 
 
-@dataclass
-class _AppStateTracker:
-    interrupt: bool = False
-    is_connected: bool = False
-
-    def register_signal_handler(self) -> None:
-        """Register handlers for exit signals."""
-
-        def signal_handler(sig, frame):  # type: ignore
-            # pylint: disable=unused-argument
-            self.interrupt = True
-            raise StopIteration from None
-
-        signal.signal(signal.SIGINT, signal_handler)
-        signal.signal(signal.SIGTERM, signal_handler)
+def _run_flwr_clientapp(args: list[str], main_pid: int) -> None:
+    # Monitor the main process in case of SIGKILL
+    def main_process_monitor() -> None:
+        while True:
+            time.sleep(1)
+            if os.getppid() != main_pid:
+                os.kill(os.getpid(), 9)
 
+    threading.Thread(target=main_process_monitor, daemon=True).start()
 
-def _run_flwr_clientapp(args: list[str]) -> None:
+    # Run the command
     sys.argv = args
     flwr_clientapp()
 

{flwr_nightly-1.15.0.dev20250115 → flwr_nightly-1.15.0.dev20250117}/src/py/flwr/client/grpc_client/connection.py

@@ -47,12 +47,6 @@ from flwr.proto.transport_pb2 import (  # pylint: disable=E0611
 )
 from flwr.proto.transport_pb2_grpc import FlowerServiceStub  # pylint: disable=E0611
 
-# The following flags can be uncommented for debugging. Other possible values:
-# https://github.com/grpc/grpc/blob/master/doc/environment_variables.md
-# import os
-# os.environ["GRPC_VERBOSITY"] = "debug"
-# os.environ["GRPC_TRACE"] = "tcp,http"
-
 
 def on_channel_state_change(channel_connectivity: str) -> None:
     """Log channel connectivity."""

{flwr_nightly-1.15.0.dev20250115 → flwr_nightly-1.15.0.dev20250117}/src/py/flwr/client/grpc_rere_client/connection.py

@@ -311,3 +311,13 @@ def grpc_request_response(  # pylint: disable=R0913,R0914,R0915,R0917
         yield (receive, send, create_node, delete_node, get_run, get_fab)
     except Exception as exc:  # pylint: disable=broad-except
         log(ERROR, exc)
+    # Cleanup
+    finally:
+        try:
+            if node is not None:
+                # Disable retrying
+                retry_invoker.max_tries = 1
+                delete_node()
+        except grpc.RpcError:
+            pass
+        channel.close()

{flwr_nightly-1.15.0.dev20250115 → flwr_nightly-1.15.0.dev20250117}/src/py/flwr/client/rest_client/connection.py

@@ -26,6 +26,7 @@ from typing import Callable, Optional, TypeVar, Union
 
 from cryptography.hazmat.primitives.asymmetric import ec
 from google.protobuf.message import Message as GrpcMessage
+from requests.exceptions import ConnectionError as RequestsConnectionError
 
 from flwr.client.heartbeat import start_ping_loop
 from flwr.client.message_handler.message_handler import validate_out_message
@@ -379,3 +380,12 @@ def http_request_response(  # pylint: disable=R0913,R0914,R0915,R0917
         yield (receive, send, create_node, delete_node, get_run, get_fab)
     except Exception as exc:  # pylint: disable=broad-except
         log(ERROR, exc)
+    # Cleanup
+    finally:
+        try:
+            if node is not None:
+                # Disable retrying
+                retry_invoker.max_tries = 1
+                delete_node()
+        except RequestsConnectionError:
+            pass

{flwr_nightly-1.15.0.dev20250115 → flwr_nightly-1.15.0.dev20250117}/src/py/flwr/client/supernode/app.py

@@ -86,6 +86,11 @@ def run_supernode() -> None:
 
     log(DEBUG, "Isolation mode: %s", args.isolation)
 
+    # Register handlers for graceful shutdown
+    register_exit_handlers(
+        event_type=EventType.RUN_SUPERNODE_LEAVE,
+    )
+
     start_client_internal(
         server_address=args.superlink,
         load_client_app_fn=load_fn,
@@ -103,11 +108,6 @@ def run_supernode() -> None:
         clientappio_api_address=args.clientappio_api_address,
     )
 
-    # Graceful shutdown
-    register_exit_handlers(
-        event_type=EventType.RUN_SUPERNODE_LEAVE,
-    )
-
 
 def run_client_app() -> None:
     """Run Flower client app."""

{flwr_nightly-1.15.0.dev20250115 → flwr_nightly-1.15.0.dev20250117}/src/py/flwr/common/grpc.py

@@ -16,6 +16,7 @@
 
 
 import concurrent.futures
+import os
 import sys
 from collections.abc import Sequence
 from logging import DEBUG, ERROR
@@ -35,6 +36,12 @@ INVALID_CERTIFICATES_ERR_MSG = """
 
 AddServicerToServerFn = Callable[..., Any]
 
+if "GRPC_VERBOSITY" not in os.environ:
+    os.environ["GRPC_VERBOSITY"] = "error"
+# The following flags can be uncommented for debugging. Other possible values:
+# https://github.com/grpc/grpc/blob/master/doc/environment_variables.md
+# os.environ["GRPC_TRACE"] = "tcp,http"
+
 
 def create_channel(
     server_address: str,

{flwr_nightly-1.15.0.dev20250115 → flwr_nightly-1.15.0.dev20250117}/src/py/flwr/server/app.py

@@ -20,6 +20,7 @@ import csv
 import importlib.util
 import multiprocessing
 import multiprocessing.context
+import os
 import sys
 import threading
 from collections.abc import Sequence
@@ -447,7 +448,17 @@ def run_superlink() -> None:
     sys.exit(1)
 
 
-def _run_flwr_command(args: list[str]) -> None:
+def _run_flwr_command(args: list[str], main_pid: int) -> None:
+    # Monitor the main process in case of SIGKILL
+    def main_process_monitor() -> None:
+        while True:
+            sleep(1)
+            if os.getppid() != main_pid:
+                os.kill(os.getpid(), 9)
+
+    threading.Thread(target=main_process_monitor, daemon=True).start()
+
+    # Run the command
     sys.argv = args
     if args[0] == "flwr-serverapp":
         flwr_serverapp()
@@ -493,7 +504,7 @@ def _flwr_scheduler(
             ]
 
             proc = mp_spawn_context.Process(
-                target=_run_flwr_command, args=(command,), daemon=True
+                target=_run_flwr_command, args=(command, os.getpid()), daemon=True
             )
             proc.start()
 

{flwr_nightly-1.15.0.dev20250115 → flwr_nightly-1.15.0.dev20250117}/src/py/flwr/server/serverapp/app.py

@@ -117,6 +117,7 @@ def run_serverapp(  # pylint: disable=R0914, disable=W0212, disable=R0915
     log_uploader = None
     success = True
     hash_run_id = None
+    run_status = None
     while True:
 
         try: