feat(boto3): Use object.* ops for S3 Object operations (#8037)

### Description
S3 object operations now override default `rpc` op with `object.*`.
Other S3 operations, e.g. `ListBuckets`, stay on `rpc`.

Author: pabloDeputter

sentry_sdk/consts.py

@@ -1477,6 +1477,16 @@ class OP:
     HUGGINGFACE_HUB_CHAT_COMPLETIONS_CREATE = (
         "ai.chat_completions.create.huggingface_hub"
     )
+    OBJECT_COPY = "object.copy"
+    OBJECT_DELETE = "object.delete"
+    OBJECT_GET = "object.get"
+    OBJECT_HEAD = "object.head"
+    OBJECT_LIST = "object.list"
+    OBJECT_MULTIPART_UPLOAD_ABORT = "object.multipart_upload.abort"
+    OBJECT_MULTIPART_UPLOAD_COMPLETE = "object.multipart_upload.complete"
+    OBJECT_MULTIPART_UPLOAD_CREATE = "object.multipart_upload.create"
+    OBJECT_PUT = "object.put"
+    OBJECT_UPLOAD_PART = "object.upload_part"
     QUEUE_PROCESS = "queue.process"
     QUEUE_PUBLISH = "queue.publish"
     QUEUE_SUBMIT_ARQ = "queue.submit.arq"

sentry_sdk/integrations/boto3/_services/s3.py

@@ -2,7 +2,7 @@
 from datetime import datetime
 from typing import TYPE_CHECKING
 
-from sentry_sdk.consts import SPANDATA
+from sentry_sdk.consts import OP, SPANDATA
 from sentry_sdk.integrations.boto3._services.base import _ServiceExtension
 from sentry_sdk.integrations.boto3._utils import (
     _extract_attributes,
@@ -11,14 +11,36 @@
 from sentry_sdk.utils import capture_internal_exceptions
 
 if TYPE_CHECKING:
-    from typing import Any, Sequence
+    from typing import Any, Dict, Optional, Sequence
 
     from sentry_sdk._types import Attributes
     from sentry_sdk.integrations.boto3._context import AwsCallContext
     from sentry_sdk.integrations.boto3._utils import (
         _AttributeSpec,
     )
 
+# s3 object apis get `object.*` ops, every other s3 api, e.g. `ListBuckets`, keeps the default `rpc`.
+# https://github.com/getsentry/sentry-conventions/blob/main/model/op/object.json
+_OBJECT_OPS: "Dict[str, str]" = {
+    "GetObject": OP.OBJECT_GET,
+    "HeadObject": OP.OBJECT_HEAD,
+    # `GetObjectAttributes` combines `HeadObject` and `ListParts`.
+    # https://docs.aws.amazon.com/AmazonS3/latest/API/API_GetObjectAttributes.html
+    "GetObjectAttributes": OP.OBJECT_HEAD,
+    "PutObject": OP.OBJECT_PUT,
+    "CopyObject": OP.OBJECT_COPY,
+    "DeleteObject": OP.OBJECT_DELETE,
+    "DeleteObjects": OP.OBJECT_DELETE,
+    "ListObjects": OP.OBJECT_LIST,
+    "ListObjectsV2": OP.OBJECT_LIST,
+    "ListObjectVersions": OP.OBJECT_LIST,
+    "UploadPart": OP.OBJECT_UPLOAD_PART,
+    "UploadPartCopy": OP.OBJECT_UPLOAD_PART,
+    "CreateMultipartUpload": OP.OBJECT_MULTIPART_UPLOAD_CREATE,
+    "CompleteMultipartUpload": OP.OBJECT_MULTIPART_UPLOAD_COMPLETE,
+    "AbortMultipartUpload": OP.OBJECT_MULTIPART_UPLOAD_ABORT,
+}
+
 # maps s3 operations to the response field that contains the complete file size.
 _RESPONSE_FILE_SIZE_FIELDS = {
     "GetObjectAttributes": "ObjectSize",
@@ -36,6 +58,9 @@
 class _S3Extension(_ServiceExtension):
     __slots__ = ()
 
+    def get_span_op(self, ctx: "AwsCallContext") -> "Optional[str]":
+        return _OBJECT_OPS.get(ctx.operation_name)
+
     def get_request_attributes(self, ctx: "AwsCallContext") -> "Attributes":
         attributes: "Attributes" = _extract_attributes(ctx.params, _REQUEST_ATTRIBUTES)
         with capture_internal_exceptions():

tests/integrations/boto3/test_client.py

@@ -16,6 +16,7 @@
 from sentry_sdk.integrations.boto3 import Boto3Integration
 from sentry_sdk.integrations.boto3._services.base import _ServiceExtension
 from sentry_sdk.integrations.boto3._services.registry import _SERVICE_EXTENSIONS
+from sentry_sdk.integrations.boto3._services.s3 import _S3Extension
 from sentry_sdk.integrations.boto3.consts import (
     AWS_RPC_SYSTEM_NAME,
     CLOUD_PROVIDER,
@@ -146,7 +147,7 @@ def record_client_span(request, **kwargs):
         if span["name"] == "S3.GetObject"
         and (
             span.get("attributes", {}).get(SPANDATA.SENTRY_ORIGIN) == ORIGIN
-            and span.get("attributes", {}).get(SPANDATA.SENTRY_OP) == OP.RPC
+            and span.get("attributes", {}).get(SPANDATA.SENTRY_OP) == OP.OBJECT_GET
         )
     ]
     http_spans = [
@@ -251,7 +252,7 @@ def test_omit_url_data_if_parsing_fails(capture_items, client_factory):
             )
 
     parse_url.assert_called()
-    (span,) = spans_by_op[OP.RPC]
+    (span,) = spans_by_op[OP.OBJECT_HEAD]
     attributes = span.get("attributes", {})
     assert SPANDATA.URL_FULL not in attributes
     assert SPANDATA.URL_FRAGMENT not in attributes
@@ -326,7 +327,7 @@ def test_url_query_data_collection(
     (span,) = [
         item.payload
         for item in items
-        if item.payload.get("attributes", {}).get(SPANDATA.SENTRY_OP) == OP.RPC
+        if item.payload.get("attributes", {}).get(SPANDATA.SENTRY_OP) == OP.OBJECT_LIST
     ]
     attributes = span.get("attributes", {})
 
@@ -428,9 +429,8 @@ def _capture_stubbed_client_span(
         )
         stubber.assert_no_pending_responses()
 
-    client_spans = spans_by_op.get(OP.RPC, [])
-    assert len(client_spans) == 1
-    return client_spans[0]
+    ((client_span,),) = spans_by_op.values()
+    return client_span
 
 
 def test_service_extension_customizes_client_span(
@@ -503,6 +503,7 @@ def get_response_attributes(self, ctx, response):
         "method_name",
         "api_params",
         "span_name",
+        "expected_op",
         "rpc_service",
         "rpc_method",
         "endpoint_url",
@@ -515,6 +516,7 @@ def get_response_attributes(self, ctx, response):
             "head_object",
             {"Bucket": "bucket", "Key": "foo"},
             "S3.HeadObject",
+            OP.OBJECT_HEAD,
             "S3",
             "HeadObject",
             "http://localhost:4566",
@@ -526,6 +528,7 @@ def get_response_attributes(self, ctx, response):
             "list_event_buses",
             {},
             "EventBridge.ListEventBuses",
+            OP.RPC,
             "EventBridge",
             "ListEventBuses",
             None,
@@ -537,6 +540,7 @@ def get_response_attributes(self, ctx, response):
             "get_rest_apis",
             {},
             "APIGateway.GetRestApis",
+            OP.RPC,
             "APIGateway",
             "GetRestApis",
             None,
@@ -552,6 +556,7 @@ def test_client_call_has_common_attributes(
     method_name,
     api_params,
     span_name,
+    expected_op,
     rpc_service,
     rpc_method,
     endpoint_url,
@@ -576,7 +581,7 @@ def test_client_call_has_common_attributes(
 
     assert span["name"] == span_name
     assert span.get("end_timestamp") is not None
-    assert attributes[SPANDATA.SENTRY_OP] == OP.RPC
+    assert attributes[SPANDATA.SENTRY_OP] == expected_op
     assert attributes[SPANDATA.SENTRY_ORIGIN] == ORIGIN
     assert attributes[SPANDATA.SENTRY_KIND] == "client"
     assert attributes[SPANDATA.CLOUD_PROVIDER] == CLOUD_PROVIDER
@@ -647,7 +652,7 @@ def test_client_call_has_response_header_attributes(
             capture_items,
         )
 
-    spans = spans_by_op[OP.RPC]
+    spans = spans_by_op[OP.OBJECT_HEAD]
     assert len(spans) == 1
     attributes = spans[0].get("attributes", {})
     assert attributes[SPANDATA.HTTP_STATUS_CODE] == 200
@@ -666,7 +671,7 @@ def test_retry_attempts_share_one_client_span(
     spans_by_op = _capture_boto3_spans_by_op(
         lambda: client.head_object(Bucket="bucket", Key="foo"), capture_items
     )
-    client_spans = spans_by_op.get(OP.RPC, [])
+    client_spans = spans_by_op.get(OP.OBJECT_HEAD, [])
 
     assert len(request_span_ids) == attempt_count
     # all `AWSRequest` instances created during retries reference the same client span.
@@ -689,7 +694,7 @@ def attempt_failed_head_object_call():
     spans_by_op = _capture_boto3_spans_by_op(
         attempt_failed_head_object_call, capture_items
     )
-    client_spans = spans_by_op.get(OP.RPC, [])
+    client_spans = spans_by_op.get(OP.OBJECT_HEAD, [])
 
     assert len(request_span_ids) == 2
     assert len(set(request_span_ids)) == 1
@@ -706,7 +711,7 @@ def test_client_error_has_response_attributes_and_is_unchanged(
     monkeypatch,
     with_service_extension,
 ):
-    class TestServiceExtension(_ServiceExtension):
+    class TestServiceExtension(_S3Extension):
         def get_response_attributes(self, ctx, response):
             return {
                 "aws.test.error": response["Error"]["Code"],
@@ -745,14 +750,14 @@ def invoke_failing_client_method():
     spans_by_op = _capture_boto3_spans_by_op(
         invoke_failing_client_method, capture_items
     )
-    client_spans = spans_by_op.get(OP.RPC, [])
+    client_spans = spans_by_op.get(OP.OBJECT_HEAD, [])
     _assert_one_failed_span(client_spans)
     attributes = client_spans[0].get("attributes", {})
 
     span = client_spans[0]
     assert span["name"] == "S3.HeadObject"
     assert span.get("end_timestamp") is not None
-    assert attributes[SPANDATA.SENTRY_OP] == OP.RPC
+    assert attributes[SPANDATA.SENTRY_OP] == OP.OBJECT_HEAD
     assert attributes[SPANDATA.SENTRY_ORIGIN] == ORIGIN
     assert attributes[SPANDATA.SENTRY_KIND] == "client"
     assert attributes[SPANDATA.CLOUD_PROVIDER] == CLOUD_PROVIDER
@@ -806,7 +811,7 @@ def invoke_failing_client_method():
     spans_by_op = _capture_boto3_spans_by_op(
         invoke_failing_client_method, capture_items
     )
-    client_spans = spans_by_op.get(OP.RPC, [])
+    client_spans = spans_by_op.get(OP.OBJECT_HEAD, [])
     _assert_one_failed_span(client_spans)
 
     expected_error_type = (
@@ -852,7 +857,7 @@ def invoke_client_method():
         stubber.add_response("head_object", original_response, api_params)
         spans_by_op = _capture_boto3_spans_by_op(invoke_client_method, capture_items)
 
-    client_spans = spans_by_op.get(OP.RPC, [])
+    client_spans = spans_by_op.get(OP.OBJECT_HEAD, [])
     assert returned_responses == [original_response]
     assert returned_responses[0] is original_response
     if failing_instrumentation == "_get_response_attributes":
@@ -889,7 +894,7 @@ def invoke_failing_client_method():
     spans_by_op = _capture_boto3_spans_by_op(
         invoke_failing_client_method, capture_items
     )
-    client_spans = spans_by_op.get(OP.RPC, [])
+    client_spans = spans_by_op.get(OP.OBJECT_HEAD, [])
 
     assert len(client_spans) == 1
     assert client_spans[0]["status"] == "error"
@@ -922,7 +927,7 @@ def invoke_client_method_and_read_body():
     spans_by_op = _capture_boto3_spans_by_op(
         invoke_client_method_and_read_body, capture_items
     )
-    client_spans = spans_by_op.get(OP.RPC, [])
+    client_spans = spans_by_op.get(OP.OBJECT_GET, [])
     stream_spans = spans_by_op.get(OP.HTTP_CLIENT_STREAM, [])
 
     assert len(client_spans) == 1
@@ -971,7 +976,7 @@ def invoke_client_method_and_read_body():
     spans_by_op = _capture_boto3_spans_by_op(
         invoke_client_method_and_read_body, capture_items
     )
-    client_spans = spans_by_op.get(OP.RPC, [])
+    client_spans = spans_by_op.get(OP.OBJECT_GET, [])
     stream_spans = spans_by_op.get(OP.HTTP_CLIENT_STREAM, [])
 
     assert len(client_spans) == 1

tests/integrations/boto3/test_s3.py

@@ -5,7 +5,7 @@
 import pytest
 from botocore.stub import Stubber
 
-from sentry_sdk.consts import OP, SPANDATA
+from sentry_sdk.consts import SPANDATA
 from sentry_sdk.integrations.boto3.consts import AWS_RPC_SYSTEM_NAME, ORIGIN
 from tests.integrations.boto3.helpers import (
     capture_spans_by_op,
@@ -26,12 +26,11 @@ def _stubbed_span(client, capture_items, method, params, response):
             lambda: getattr(client, method)(**params), capture_items
         )
         stubber.assert_no_pending_responses()
-    (span,) = spans[OP.RPC]
+    ((span,),) = spans.values()
     operation = client.meta.method_to_api_mapping[method]
     assert span["name"] == "S3.%s" % operation
     assert span.get("end_timestamp") is not None
     attributes = span.get("attributes", {})
-    assert attributes[SPANDATA.SENTRY_OP] == OP.RPC
     assert attributes[SPANDATA.SENTRY_ORIGIN] == ORIGIN
     assert attributes[SPANDATA.SENTRY_KIND] == "client"
     assert attributes[SPANDATA.CLOUD_PROVIDER] == "aws"