test(serve): stop unit tests resolving IAM roles over the network

sagemaker-core/tests/unit/helper/test_iam_role_resolver.py

@@ -24,6 +24,8 @@
     _role_trusts_service,
     _trusted_services_in_document,
     _expected_trust_services,
+    _config_path_for_role_type,
+    _resolve_config_default_role,
 )
 
 
@@ -1391,3 +1393,27 @@ class TestBackwardCompatibleExceptions:
 
     def test_role_auto_creation_error_importable(self):
         assert issubclass(RoleAutoCreationError, Exception)
+
+
+class TestConfigDefaultRole:
+    """Config-default lookup is only consulted for role types that have a path."""
+
+    @pytest.mark.parametrize("role_type", ["training", "feature_store"])
+    def test_role_types_with_a_config_path(self, role_type):
+        assert _config_path_for_role_type(role_type) is not None
+
+    @pytest.mark.parametrize("role_type", ["serving", "bedrock", "not_a_role_type"])
+    def test_role_types_without_a_config_path(self, role_type):
+        assert _config_path_for_role_type(role_type) is None
+
+    @pytest.mark.parametrize("role_type", ["serving", "bedrock"])
+    def test_returns_none_without_reading_config(self, role_type):
+        """No config path means no config to read, so resolution short-circuits.
+
+        The caller then falls back to caller-identity resolution. Asserting that
+        resolve_value_from_config is never consulted keeps the short-circuit from
+        regressing into a config read for role types that have no default.
+        """
+        with patch("sagemaker.core.common_utils.resolve_value_from_config") as mock_resolve:
+            assert _resolve_config_default_role(role_type) is None
+        mock_resolve.assert_not_called()

sagemaker-serve/tests/unit/conftest.py

@@ -0,0 +1,71 @@
+# Copyright Amazon.com, Inc. or its affiliates. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License"). You
+# may not use this file except in compliance with the License. A copy of
+# the License is located at
+#
+#     http://aws.amazon.com/apache2.0/
+#
+# or in the "license" file accompanying this file. This file is
+# distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF
+# ANY KIND, either express or implied. See the License for the specific
+# language governing permissions and limitations under the License.
+"""Shared fixtures for the sagemaker-serve unit tests.
+
+Keeps role resolution off the network. ``ModelBuilder.__post_init__`` resolves a
+default execution role when none is supplied, and ``resolve_and_validate_role``
+does that by calling ``sts:GetCallerIdentity`` followed by a paginated
+``iam:SimulatePrincipalPolicy``. Any unit test that builds a ModelBuilder without
+passing ``role_arn`` therefore makes live IAM calls, which only succeed because
+CI credentials happen to be present.
+
+Those calls are the dominant source of CI flakiness for this package: IAM
+throttles on a shared, account-wide quota, so builds fail with
+``Throttling ... SimulatePrincipalPolicy ... reached max retries`` on whichever
+test happened to be running. The failures are not reproducible -- each build runs
+the suite three times (py310/py311/py312) and these failures almost always hit
+only one of the three.
+"""
+from __future__ import annotations
+
+from contextlib import ExitStack
+from unittest.mock import patch
+
+import pytest
+
+# Stands in for "the caller's own identity role", which the real resolver obtains
+# from sts:GetCallerIdentity.
+DEFAULT_UNIT_TEST_ROLE = "arn:aws:iam::123456789012:role/UnitTestExecutionRole"
+
+# resolve_and_validate_role has to be patched once per importing module, not once
+# at its definition site: both modules below bind it with a module-level
+# ``from ... import resolve_and_validate_role``, so each holds its own reference
+# and patching sagemaker.core.helper.iam_role_resolver would not rebind either.
+_RESOLVER_LOOKUP_SITES = (
+    "sagemaker.serve.model_builder.resolve_and_validate_role",
+    "sagemaker.serve.bedrock_model_builder.resolve_and_validate_role",
+)
+
+
+def _resolve_role_offline(provided_role=None, *, role_arn=None, **kwargs):
+    """Mirror resolve_and_validate_role's contract without touching IAM.
+
+    The real function returns the role it was given and only falls back to the
+    caller's identity when none was supplied. Preserving that passthrough matters:
+    four of its five call sites in this package pass a real role, and tests assert
+    on the value that comes back.
+    """
+    return provided_role or role_arn or DEFAULT_UNIT_TEST_ROLE
+
+
+@pytest.fixture(autouse=True)
+def stub_iam_role_resolution():
+    """Resolve execution roles locally for every unit test in this package.
+
+    Tests that patch ``resolve_and_validate_role`` themselves still win, because
+    their patch is applied inside this one and so rebinds the same attribute last.
+    """
+    with ExitStack() as stack:
+        for target in _RESOLVER_LOOKUP_SITES:
+            stack.enter_context(patch(target, side_effect=_resolve_role_offline))
+        yield