feat(storage): Enable full object checksum PR 3/3: integrate full-object checksum in AsyncMultiRangeDownloader

packages/google-cloud-storage/google/cloud/storage/asyncio/async_multi_range_downloader.py

@@ -45,6 +45,7 @@
     _ReadResumptionStrategy,
 )
 
+from google.cloud.storage.exceptions import DataCorruption
 from ._utils import raise_if_no_fast_crc32c
 
 _MAX_READ_RANGES_PER_BIDI_READ_REQUEST = 100
@@ -219,8 +220,6 @@ def __init__(
             )
             generation = kwargs.pop("generation_number")
 
-        raise_if_no_fast_crc32c()
-
         self.client = client
         self.bucket_name = bucket_name
         self.object_name = object_name
@@ -232,6 +231,8 @@ def __init__(
         self._multiplexer: Optional[_StreamMultiplexer] = None
         self.persisted_size: Optional[int] = None  # updated after opening the stream
         self._open_retries: int = 0
+        self.is_finalized: bool = False
+        self.full_obj_server_crc32c: Optional[int] = None
 
     async def __aenter__(self):
         """Opens the underlying bidi-gRPC connection to read from the object."""
@@ -327,6 +328,8 @@ async def _do_open():
                 self.read_handle = self.read_obj_str.read_handle
             if self.read_obj_str.persisted_size is not None:
                 self.persisted_size = self.read_obj_str.persisted_size
+            self.is_finalized = self.read_obj_str.is_finalized
+            self.full_obj_server_crc32c = self.read_obj_str.full_obj_server_crc32c
 
             self._is_stream_open = True
 
@@ -363,6 +366,8 @@ async def factory():
                 self.generation = stream.generation_number
             if stream.read_handle:
                 self.read_handle = stream.read_handle
+            self.is_finalized = stream.is_finalized
+            self.full_obj_server_crc32c = stream.full_obj_server_crc32c
 
             self.read_obj_str = stream
             self._is_stream_open = True
@@ -377,6 +382,7 @@ async def download_ranges(
         lock: asyncio.Lock = None,
         retry_policy: Optional[AsyncRetry] = None,
         metadata: Optional[List[Tuple[str, str]]] = None,
+        enable_checksum: bool = True,
     ) -> None:
         """Downloads multiple byte ranges from the object into the buffers
         provided by user with automatic retries.
@@ -412,6 +418,9 @@ async def download_ranges(
                 "Invalid input - length of read_ranges cannot be more than 1000"
             )
 
+        if enable_checksum:
+            raise_if_no_fast_crc32c()
+
         if not self._is_stream_open:
             raise ValueError("Underlying bidi-gRPC stream is not open")
 
@@ -422,16 +431,29 @@ async def download_ranges(
         download_states = {}
         for read_range in read_ranges:
             read_id = generate_random_56_bit_integer()
+            # Unpack tuple into self-documenting variable names to improve readability.
+            offset, length, user_buffer = read_range
+
+            # Heuristic to detect full object reads:
+            # - Implicit full object read: start offset is 0 and length is 0 (read all).
+            # - Explicit full object read: start offset is 0 and length matches the exact persisted size.
+            is_full_object_read = (
+                (offset == 0 and length == 0) or
+                (self.persisted_size is not None and offset == 0 and length == self.persisted_size)
+            )
             download_states[read_id] = _DownloadState(
-                initial_offset=read_range[0],
-                initial_length=read_range[1],
-                user_buffer=read_range[2],
+                initial_offset=offset,
+                initial_length=length,
+                user_buffer=user_buffer,
+                is_full_object_read=is_full_object_read,
             )
 
         initial_state = {
             "download_states": download_states,
             "read_handle": self.read_handle,
             "routing_token": None,
+            "enable_checksum": enable_checksum,
+            "full_obj_server_crc32c": self.full_obj_server_crc32c,
         }
 
         read_ids = set(download_states.keys())
@@ -519,12 +541,18 @@ async def generator():
                 strategy, send_and_recv_via_multiplexer
             )
 
-            await retry_manager.execute(initial_state, retry_policy)
+            try:
+                await retry_manager.execute(initial_state, retry_policy)
+            except DataCorruption:
+                if self.is_stream_open:
+                    await self.close()
+                raise
 
             if initial_state.get("read_handle"):
                 self.read_handle = initial_state["read_handle"]
         finally:
-            self._multiplexer.unregister(read_ids)
+            if self._multiplexer is not None:
+                self._multiplexer.unregister(read_ids)
 
     async def close(self):
         """

packages/google-cloud-storage/google/cloud/storage/asyncio/async_read_object_stream.py

@@ -79,6 +79,9 @@ def __init__(
         self.socket_like_rpc: Optional[AsyncBidiRpc] = None
         self._is_stream_open: bool = False
         self.persisted_size: Optional[int] = None
+        self.is_finalized: bool = False
+        self.full_obj_server_crc32c: Optional[int] = None
+        self.object_metadata: Optional[_storage_v2.Object] = None
 
     async def open(self, metadata: Optional[List[Tuple[str, str]]] = None) -> None:
         """Opens the bidi-gRPC connection to read from the object.
@@ -132,6 +135,18 @@ async def open(self, metadata: Optional[List[Tuple[str, str]]] = None) -> None:
                 self.generation_number = response.metadata.generation
             # update persisted size
             self.persisted_size = response.metadata.size
+            self.object_metadata = response.metadata
+            if (
+                hasattr(response.metadata, "finalize_time")
+                and response.metadata.finalize_time
+                and response.metadata.finalize_time.second > 0
+            ):
+                self.is_finalized = True
+                if (
+                    hasattr(response.metadata, "checksums")
+                    and response.metadata.checksums
+                ):
+                    self.full_obj_server_crc32c = response.metadata.checksums.crc32c
 
         if response and response.read_handle:
             self.read_handle = response.read_handle

packages/google-cloud-storage/google/cloud/storage/asyncio/retry/reads_resumption_strategy.py

@@ -36,14 +36,22 @@ class _DownloadState:
     """A helper class to track the state of a single range download."""
 
     def __init__(
-        self, initial_offset: int, initial_length: int, user_buffer: IO[bytes]
+        self,
+        initial_offset: int,
+        initial_length: int,
+        user_buffer: IO[bytes],
+        is_full_object_read: bool = False,
     ):
         self.initial_offset = initial_offset
         self.initial_length = initial_length
         self.user_buffer = user_buffer
         self.bytes_written = 0
         self.next_expected_offset = initial_offset
         self.is_complete = False
+        self.is_full_object_read = is_full_object_read
+        self.rolling_checksum = (
+            google_crc32c.Checksum() if is_full_object_read else None
+        )
 
 
 class _ReadResumptionStrategy(_BaseResumptionStrategy):
@@ -90,6 +98,7 @@ def update_state_from_response(
             )
 
         download_states = state["download_states"]
+        checksum_enabled = state.get("enable_checksum", True)
 
         for object_data_range in proto.object_data_ranges:
             # Ignore empty ranges or ranges for IDs not in our state
@@ -125,7 +134,7 @@ def update_state_from_response(
             checksummed_data = object_data_range.checksummed_data
             data = checksummed_data.content
 
-            if checksummed_data.HasField("crc32c"):
+            if checksum_enabled and checksummed_data.HasField("crc32c"):
                 server_checksum = checksummed_data.crc32c
                 client_checksum = google_crc32c.value(data)
                 if server_checksum != client_checksum:
@@ -138,10 +147,14 @@ def update_state_from_response(
             # Update State & Write Data
             chunk_size = len(data)
             read_state.user_buffer.write(data)
+
+            # Commit updates only after the write succeeds
+            if checksum_enabled and read_state.rolling_checksum is not None:
+                read_state.rolling_checksum.update(data)
             read_state.bytes_written += chunk_size
             read_state.next_expected_offset += chunk_size
 
-            # Final Byte Count Verification
+            # Final Byte Count & Full Object Checksum Verification
             if object_data_range.range_end:
                 read_state.is_complete = True
                 if (
@@ -154,6 +167,22 @@ def update_state_from_response(
                         f"Expected {read_state.initial_length}, got {read_state.bytes_written}",
                     )
 
+                # Perform full-object checksum verification once the stream finishes.
+                if read_state.is_full_object_read and checksum_enabled:
+                    full_obj_server_crc32c = state.get("full_obj_server_crc32c")
+                    if full_obj_server_crc32c is not None:
+                        # Use standard big-endian byte conversion to retrieve the rolling checksum value.
+                        client_checksum = int.from_bytes(
+                            read_state.rolling_checksum.digest(),
+                            byteorder="big",
+                        )
+                        if client_checksum != full_obj_server_crc32c:
+                            raise DataCorruption(
+                                response,
+                                f"Full object checksum mismatch for read_id {read_id}. "
+                                f"Server authoritative crc32c: {full_obj_server_crc32c}, client calculated rolling: {client_checksum}.",
+                            )
+
     async def recover_state_on_failure(self, error: Exception, state: Any) -> None:
         """Handles BidiReadObjectRedirectedError for reads."""
         routing_token, read_handle = _handle_redirect(error)

packages/google-cloud-storage/tests/system/test_zonal.py

@@ -26,6 +26,7 @@
     ObjectContexts,
     ObjectCustomContextPayload,
 )
+from google.cloud.storage.exceptions import DataCorruption
 
 pytestmark = pytest.mark.skipif(
     os.getenv("RUN_ZONAL_SYSTEM_TESTS") != "True",
@@ -961,3 +962,80 @@ async def _run():
         blobs_to_delete.append(storage_client.bucket(_ZONAL_BUCKET).blob(object_name))
 
     event_loop.run_until_complete(_run())
+
+
+@pytest.mark.parametrize(
+    "read_start, read_length, enable_checksum",
+    [
+        (0, 0, True),
+        (0, 1024 * 1024, True),
+        (0, 0, False),
+    ],
+)
+def test_mrd_checksum_validation(
+    storage_client, blobs_to_delete, event_loop, grpc_client_direct, read_start, read_length, enable_checksum
+):
+    """
+    Tests full downloads with specified offset, length, and enable_checksum toggle on finalized objects.
+    """
+    object_size = 1024 * 1024  # 1MB
+    object_name = f"test_mrd_chksum-{uuid.uuid4()}"
+
+    async def _run():
+        object_data = os.urandom(object_size)
+
+        writer = AsyncAppendableObjectWriter(
+            grpc_client_direct, _ZONAL_BUCKET, object_name
+        )
+        await writer.open()
+        await writer.append(object_data)
+        await writer.close(finalize_on_close=True)
+
+        async with AsyncMultiRangeDownloader(
+            grpc_client_direct, _ZONAL_BUCKET, object_name
+        ) as mrd:
+            buffer = BytesIO()
+            await mrd.download_ranges([(read_start, read_length, buffer)], enable_checksum=enable_checksum)
+            assert buffer.getvalue() == object_data
+
+        # cleanup
+        del writer
+        gc.collect()
+        blobs_to_delete.append(storage_client.bucket(_ZONAL_BUCKET).blob(object_name))
+
+    event_loop.run_until_complete(_run())
+
+
+def test_mrd_checksum_unfinalized_appendable_skipped(
+    storage_client, blobs_to_delete, event_loop, grpc_client_direct
+):
+    """
+    Verifies that live, unfinalized appendable objects skip the full-object checksum check
+    naturally without raising any exceptions.
+    """
+    object_name = f"test_mrd_chksum_unfin-{uuid.uuid4()}"
+
+    async def _run():
+        writer = AsyncAppendableObjectWriter(
+            grpc_client_direct, _ZONAL_BUCKET, object_name
+        )
+        await writer.open()
+        await writer.append(_BYTES_TO_UPLOAD)
+        await writer.flush()  # Flushed but not finalized!
+
+        # Download the unfinalized appendable object with enable_checksum=True
+        async with AsyncMultiRangeDownloader(
+            grpc_client_direct, _ZONAL_BUCKET, object_name
+        ) as mrd:
+            buffer = BytesIO()
+            # Since it's unfinalized, it should skip the checksum check without raising
+            await mrd.download_ranges([(0, 0, buffer)], enable_checksum=True)
+            assert buffer.getvalue() == _BYTES_TO_UPLOAD
+
+        # cleanup
+        await writer.close()
+        del writer
+        gc.collect()
+        blobs_to_delete.append(storage_client.bucket(_ZONAL_BUCKET).blob(object_name))
+
+    event_loop.run_until_complete(_run())