modal 1.1.1.dev6__tar.gz → 1.1.1.dev9__tar.gz

{modal-1.1.1.dev6 → modal-1.1.1.dev9}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: modal
-Version: 1.1.1.dev6
+Version: 1.1.1.dev9
 Summary: Python client library for Modal
 Author-email: Modal Labs <support@modal.com>
 License: Apache-2.0

{modal-1.1.1.dev6 → modal-1.1.1.dev9}/modal/_utils/blob_utils.py

@@ -188,16 +188,10 @@ def get_content_length(data: BinaryIO) -> int:
     return content_length - pos
 
 
-async def _measure_endpoint_latency(item: str) -> int:
-    latency_ms = 0
-    t0 = time.monotonic_ns()
-    async with ClientSessionRegistry.get_session().head(item) as _:
-        latency_ms = (time.monotonic_ns() - t0) // 1_000_000
-    return latency_ms
-
-
-async def _blob_upload_with_fallback(items, blob_ids: list[str], callback) -> tuple[str, bool, int]:
-    r2_latency_ms = 0
+async def _blob_upload_with_fallback(
+    items, blob_ids: list[str], callback, content_length: int
+) -> tuple[str, bool, int]:
+    r2_throughput_bytes_s = 0
     r2_failed = False
     for idx, (item, blob_id) in enumerate(zip(items, blob_ids)):
         # We want to default to R2 95% of the time and S3 5% of the time.
@@ -206,14 +200,13 @@ async def _blob_upload_with_fallback(items, blob_ids: list[str], callback) -> tu
             continue
         try:
             if blob_id.endswith(":r2"):
-                # measure the time it takes to contact the bucket endpoint
-                r2_latency_ms, _ = await asyncio.gather(
-                    _measure_endpoint_latency(item),
-                    callback(item),
-                )
+                t0 = time.monotonic_ns()
+                await callback(item)
+                dt_ns = time.monotonic_ns() - t0
+                r2_throughput_bytes_s = (content_length * 1_000_000_000) // max(dt_ns, 1)
             else:
                 await callback(item)
-            return blob_id, r2_failed, r2_latency_ms
+            return blob_id, r2_failed, r2_throughput_bytes_s
         except Exception as _:
             if blob_id.endswith(":r2"):
                 r2_failed = True
@@ -251,10 +244,11 @@ async def _blob_upload(
                 progress_report_cb=progress_report_cb,
             )
 
-        blob_id, r2_failed, r2_latency_ms = await _blob_upload_with_fallback(
+        blob_id, r2_failed, r2_throughput_bytes_s = await _blob_upload_with_fallback(
             resp.multiparts.items,
             resp.blob_ids,
             upload_multipart_upload,
+            content_length=content_length,
         )
     else:
         from .bytes_io_segment_payload import BytesIOSegmentPayload
@@ -271,16 +265,17 @@ async def _blob_upload(
                 content_md5_b64=upload_hashes.md5_base64,
             )
 
-        blob_id, r2_failed, r2_latency_ms = await _blob_upload_with_fallback(
+        blob_id, r2_failed, r2_throughput_bytes_s = await _blob_upload_with_fallback(
             resp.upload_urls.items,
             resp.blob_ids,
             upload_to_s3_url,
+            content_length=content_length,
         )
 
     if progress_report_cb:
         progress_report_cb(complete=True)
 
-    return blob_id, r2_failed, r2_latency_ms
+    return blob_id, r2_failed, r2_throughput_bytes_s
 
 
 async def blob_upload_with_r2_failure_info(payload: bytes, stub: ModalClientModal) -> tuple[str, bool, int]:
@@ -291,13 +286,13 @@ async def blob_upload_with_r2_failure_info(payload: bytes, stub: ModalClientModa
         logger.warning("Blob uploading string, not bytes - auto-encoding as utf8")
         payload = payload.encode("utf8")
     upload_hashes = get_upload_hashes(payload)
-    blob_id, r2_failed, r2_latency_ms = await _blob_upload(upload_hashes, payload, stub)
+    blob_id, r2_failed, r2_throughput_bytes_s = await _blob_upload(upload_hashes, payload, stub)
     dur_s = max(time.time() - t0, 0.001)  # avoid division by zero
     throughput_mib_s = (size_mib) / dur_s
     logger.debug(
         f"Uploaded large blob of size {size_mib:.2f} MiB ({throughput_mib_s:.2f} MiB/s, total {dur_s:.2f}s). {blob_id}"
     )
-    return blob_id, r2_failed, r2_latency_ms
+    return blob_id, r2_failed, r2_throughput_bytes_s
 
 
 async def blob_upload(payload: bytes, stub: ModalClientModal) -> str:

{modal-1.1.1.dev6 → modal-1.1.1.dev9}/modal/_utils/function_utils.py

@@ -563,7 +563,7 @@ async def _create_input(
     args_serialized = serialize((args, kwargs))
 
     if should_upload(len(args_serialized), max_object_size_bytes, function_call_invocation_type):
-        args_blob_id, r2_failed, r2_latency_ms = await blob_upload_with_r2_failure_info(args_serialized, stub)
+        args_blob_id, r2_failed, r2_throughput_bytes_s = await blob_upload_with_r2_failure_info(args_serialized, stub)
         return api_pb2.FunctionPutInputsItem(
             input=api_pb2.FunctionInput(
                 args_blob_id=args_blob_id,
@@ -572,7 +572,7 @@ async def _create_input(
             ),
             idx=idx,
             r2_failed=r2_failed,
-            r2_latency_ms=r2_latency_ms,
+            r2_throughput_bytes_s=r2_throughput_bytes_s,
         )
     else:
         return api_pb2.FunctionPutInputsItem(

{modal-1.1.1.dev6 → modal-1.1.1.dev9}/modal/client.pyi

@@ -33,7 +33,7 @@ class _Client:
         server_url: str,
         client_type: int,
         credentials: typing.Optional[tuple[str, str]],
-        version: str = "1.1.1.dev6",
+        version: str = "1.1.1.dev9",
     ):
         """mdmd:hidden
         The Modal client object is not intended to be instantiated directly by users.
@@ -163,7 +163,7 @@ class Client:
         server_url: str,
         client_type: int,
         credentials: typing.Optional[tuple[str, str]],
-        version: str = "1.1.1.dev6",
+        version: str = "1.1.1.dev9",
     ):
         """mdmd:hidden
         The Modal client object is not intended to be instantiated directly by users.

{modal-1.1.1.dev6 → modal-1.1.1.dev9}/modal/experimental/__init__.py

@@ -20,7 +20,7 @@ from ..cls import _Cls, _Obj
 from ..exception import InvalidError
 from ..image import DockerfileSpec, ImageBuilderVersion, _Image, _ImageRegistryConfig
 from ..secret import _Secret
-from .flash import flash_forward, flash_prometheus_autoscaler  # noqa: F401
+from .flash import flash_forward, flash_get_containers, flash_prometheus_autoscaler  # noqa: F401
 
 
 def stop_fetching_inputs():

{modal-1.1.1.dev6 → modal-1.1.1.dev9}/modal/experimental/flash.py

@@ -447,3 +447,20 @@ async def flash_prometheus_autoscaler(
     )
     await autoscaler.start()
     return autoscaler
+
+
+@synchronizer.create_blocking
+async def flash_get_containers(app_name: str, cls_name: str) -> list[dict[str, Any]]:
+    """
+    Return a list of flash containers for a deployed Flash service.
+
+    This is a highly experimental method that can break or be removed at any time without warning.
+    Do not use this method unless explicitly instructed to do so by Modal support.
+    """
+    client = await _Client.from_env()
+    fn = _Cls.from_name(app_name, cls_name)._class_service_function
+    assert fn is not None
+    await fn.hydrate(client=client)
+    req = api_pb2.FlashContainerListRequest(function_id=fn.object_id)
+    resp = await retry_transient_errors(client.stub.FlashContainerList, req)
+    return resp.containers

{modal-1.1.1.dev6 → modal-1.1.1.dev9}/modal/experimental/flash.pyi

@@ -250,3 +250,22 @@ class __flash_prometheus_autoscaler_spec(typing_extensions.Protocol):
         ...
 
 flash_prometheus_autoscaler: __flash_prometheus_autoscaler_spec
+
+class __flash_get_containers_spec(typing_extensions.Protocol):
+    def __call__(self, /, app_name: str, cls_name: str) -> list[dict[str, typing.Any]]:
+        """Return a list of flash containers for a deployed Flash service.
+
+        This is a highly experimental method that can break or be removed at any time without warning.
+        Do not use this method unless explicitly instructed to do so by Modal support.
+        """
+        ...
+
+    async def aio(self, /, app_name: str, cls_name: str) -> list[dict[str, typing.Any]]:
+        """Return a list of flash containers for a deployed Flash service.
+
+        This is a highly experimental method that can break or be removed at any time without warning.
+        Do not use this method unless explicitly instructed to do so by Modal support.
+        """
+        ...
+
+flash_get_containers: __flash_get_containers_spec

{modal-1.1.1.dev6 → modal-1.1.1.dev9}/modal.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: modal
-Version: 1.1.1.dev6
+Version: 1.1.1.dev9
 Summary: Python client library for Modal
 Author-email: Modal Labs <support@modal.com>
 License: Apache-2.0

{modal-1.1.1.dev6 → modal-1.1.1.dev9}/modal_proto/api.proto

@@ -1816,7 +1816,8 @@ message FunctionPutInputsItem {
   int32 idx = 1;
   FunctionInput input = 2;
   bool r2_failed = 3;
-  uint64 r2_latency_ms = 4;
+  reserved 4; // r2_latency_ms
+  uint64 r2_throughput_bytes_s = 5;
 }
 
 message FunctionPutInputsRequest {