refactor(framework): Stop retrying when TLS handshake failure detected

framework/py/flwr/common/args.py

@@ -89,7 +89,7 @@ def try_obtain_root_certificates(
     else:
         # Load the certificates if provided, or load the system certificates
         if root_cert_path is None:
-            log(INFO, "Using system certificates")
+            log(INFO, "Using system certificates for TLS connection")
             root_certificates = None
         elif not isfile(root_cert_path):
             log(ERROR, "Path argument `--root-certificates` does not point to a file.")

framework/py/flwr/common/retry_invoker.py

@@ -21,7 +21,7 @@
 import time
 from collections.abc import Callable, Generator, Iterable
 from dataclasses import dataclass
-from logging import INFO, WARN
+from logging import ERROR, INFO, WARN
 from typing import Any, cast
 
 import grpc
@@ -351,7 +351,12 @@ def _should_giveup_fn(e: Exception) -> bool:
         if e.code() == grpc.StatusCode.PERMISSION_DENIED:  # type: ignore
             raise RunNotRunningException
         if e.code() == grpc.StatusCode.UNAVAILABLE:  # type: ignore
-            return False
+            # Check if this is an SSL handshake failure - these should fail fast
+            details = str(e.details() if hasattr(e, "details") else "").lower()
+            if "handshake failed" in details:
+                log(ERROR, "SSL/TLS handshake error detected.")
+                return True  # Give up on SSL/TLS handshake errors
+            return False  # Retry on other UNAVAILABLE errors (network issues)
         return True
 
     def _wait(wait_time: float) -> None: