modal 1.1.4.dev16__py3-none-any.whl → 1.1.4.dev17__py3-none-any.whl

modal/client.pyi

@@ -33,7 +33,7 @@ class _Client:
         server_url: str,
         client_type: int,
         credentials: typing.Optional[tuple[str, str]],
-        version: str = "1.1.4.dev16",
+        version: str = "1.1.4.dev17",
     ):
         """mdmd:hidden
         The Modal client object is not intended to be instantiated directly by users.
@@ -164,7 +164,7 @@ class Client:
         server_url: str,
         client_type: int,
         credentials: typing.Optional[tuple[str, str]],
-        version: str = "1.1.4.dev16",
+        version: str = "1.1.4.dev17",
     ):
         """mdmd:hidden
         The Modal client object is not intended to be instantiated directly by users.

modal/experimental/flash.py

@@ -213,8 +213,8 @@ class _FlashPrometheusAutoscaler:
                 )
 
                 logger.warning(
-                    f"[Modal Flash] Scaling to {actual_target_containers} containers. Autoscaling decision "
-                    f"made in {time.time() - autoscaling_time} seconds."
+                    f"[Modal Flash] Scaling to {actual_target_containers=} containers. "
+                    f" Autoscaling decision made in {time.time() - autoscaling_time} seconds."
                 )
 
                 await self.autoscaling_decisions_dict.put(
@@ -223,9 +223,7 @@ class _FlashPrometheusAutoscaler:
                 )
                 await self.autoscaling_decisions_dict.put("current_replicas", actual_target_containers)
 
-                await self.cls.update_autoscaler(
-                    min_containers=actual_target_containers,
-                )
+                await self.cls.update_autoscaler(min_containers=actual_target_containers)
 
                 if time.time() - autoscaling_time < self.autoscaling_interval_seconds:
                     await asyncio.sleep(self.autoscaling_interval_seconds - (time.time() - autoscaling_time))
@@ -239,6 +237,8 @@ class _FlashPrometheusAutoscaler:
                 await asyncio.sleep(self.autoscaling_interval_seconds)
 
     async def _compute_target_containers(self, current_replicas: int) -> int:
+        # current_replicas is the number of live containers + cold starting containers (not yet live)
+        # containers is the number of live containers that are registered in flash dns
         containers = await self._get_all_containers()
         if len(containers) > current_replicas:
             logger.info(
@@ -271,11 +271,17 @@ class _FlashPrometheusAutoscaler:
             sum_metric += container_metrics[target_metric][0].value
             containers_with_metrics += 1
 
+        # n_containers_missing_metric is the number of unhealthy containers + number of cold starting containers
         n_containers_missing_metric = current_replicas - containers_with_metrics
+        # n_containers_unhealthy is the number of live containers that are not emitting metrics i.e. unhealthy
+        n_containers_unhealthy = len(containers) - containers_with_metrics
+
+        # Scale up assuming that every unhealthy container is at 2x the target metric value.
+        scale_up_target_metric_value = (sum_metric + n_containers_unhealthy * target_metric_value) / (
+            (containers_with_metrics + n_containers_unhealthy) or 1
+        )
 
-        # Scale up / down conservatively: Any container that is missing the metric is assumed to be at the minimum
-        # value of the metric when scaling up and the maximum value of the metric when scaling down.
-        scale_up_target_metric_value = sum_metric / (containers_with_metrics or 1)
+        # Scale down assuming that every container (including cold starting containers) are at the target metric value.
         scale_down_target_metric_value = (
             sum_metric + n_containers_missing_metric * target_metric_value
         ) / current_replicas
@@ -290,9 +296,14 @@ class _FlashPrometheusAutoscaler:
             desired_replicas = math.ceil(current_replicas * scale_down_ratio)
 
         logger.warning(
-            f"[Modal Flash] Current replicas: {current_replicas}, target metric value: {target_metric_value}, "
-            f"current sum of metric values: {sum_metric}, number of containers missing metric: "
-            f"{n_containers_missing_metric}, scale up ratio: {scale_up_ratio}, scale down ratio: {scale_down_ratio}, "
+            f"[Modal Flash] Current replicas: {current_replicas}, "
+            f"target metric value: {target_metric_value}, "
+            f"current sum of metric values: {sum_metric}, "
+            f"number of containers with metrics: {containers_with_metrics}, "
+            f"number of containers unhealthy: {n_containers_unhealthy}, "
+            f"number of containers missing metric (includes unhealthy): {n_containers_missing_metric}, "
+            f"scale up ratio: {scale_up_ratio}, "
+            f"scale down ratio: {scale_down_ratio}, "
             f"desired replicas: {desired_replicas}"
         )
 
@@ -312,9 +323,19 @@ class _FlashPrometheusAutoscaler:
             logger.warning(f"[Modal Flash] Error getting metrics from {url}: {e}")
             return None
 
+        # Read body with timeout/error handling and parse Prometheus metrics
+        try:
+            text_body = await response.text()
+        except asyncio.TimeoutError:
+            logger.warning(f"[Modal Flash] Timeout reading metrics body from {url}")
+            return None
+        except Exception as e:
+            logger.warning(f"[Modal Flash] Error reading metrics body from {url}: {e}")
+            return None
+
         # Parse the text-based Prometheus metrics format
         metrics: dict[str, list[Sample]] = defaultdict(list)
-        for family in text_string_to_metric_families(await response.text()):
+        for family in text_string_to_metric_families(text_body):
             for sample in family.samples:
                 metrics[sample.name] += [sample]
 

modal/functions.pyi

@@ -433,7 +433,7 @@ class Function(
 
     _call_generator: ___call_generator_spec[typing_extensions.Self]
 
-    class __remote_spec(typing_extensions.Protocol[ReturnType_INNER, P_INNER, SUPERSELF]):
+    class __remote_spec(typing_extensions.Protocol[P_INNER, ReturnType_INNER, SUPERSELF]):
         def __call__(self, /, *args: P_INNER.args, **kwargs: P_INNER.kwargs) -> ReturnType_INNER:
             """Calls the function remotely, executing it with the given arguments and returning the execution's result."""
             ...
@@ -442,7 +442,7 @@ class Function(
             """Calls the function remotely, executing it with the given arguments and returning the execution's result."""
             ...
 
-    remote: __remote_spec[modal._functions.ReturnType, modal._functions.P, typing_extensions.Self]
+    remote: __remote_spec[modal._functions.P, modal._functions.ReturnType, typing_extensions.Self]
 
     class __remote_gen_spec(typing_extensions.Protocol[SUPERSELF]):
         def __call__(self, /, *args, **kwargs) -> typing.Generator[typing.Any, None, None]:
@@ -469,7 +469,7 @@ class Function(
         """
         ...
 
-    class ___experimental_spawn_spec(typing_extensions.Protocol[ReturnType_INNER, P_INNER, SUPERSELF]):
+    class ___experimental_spawn_spec(typing_extensions.Protocol[P_INNER, ReturnType_INNER, SUPERSELF]):
         def __call__(self, /, *args: P_INNER.args, **kwargs: P_INNER.kwargs) -> FunctionCall[ReturnType_INNER]:
             """[Experimental] Calls the function with the given arguments, without waiting for the results.
 
@@ -493,7 +493,7 @@ class Function(
             ...
 
     _experimental_spawn: ___experimental_spawn_spec[
-        modal._functions.ReturnType, modal._functions.P, typing_extensions.Self
+        modal._functions.P, modal._functions.ReturnType, typing_extensions.Self
     ]
 
     class ___spawn_map_inner_spec(typing_extensions.Protocol[P_INNER, SUPERSELF]):
@@ -502,7 +502,7 @@ class Function(
 
     _spawn_map_inner: ___spawn_map_inner_spec[modal._functions.P, typing_extensions.Self]
 
-    class __spawn_spec(typing_extensions.Protocol[ReturnType_INNER, P_INNER, SUPERSELF]):
+    class __spawn_spec(typing_extensions.Protocol[P_INNER, ReturnType_INNER, SUPERSELF]):
         def __call__(self, /, *args: P_INNER.args, **kwargs: P_INNER.kwargs) -> FunctionCall[ReturnType_INNER]:
             """Calls the function with the given arguments, without waiting for the results.
 
@@ -523,7 +523,7 @@ class Function(
             """
             ...
 
-    spawn: __spawn_spec[modal._functions.ReturnType, modal._functions.P, typing_extensions.Self]
+    spawn: __spawn_spec[modal._functions.P, modal._functions.ReturnType, typing_extensions.Self]
 
     def get_raw_f(self) -> collections.abc.Callable[..., typing.Any]:
         """Return the inner Python object wrapped by this Modal Function."""

{modal-1.1.4.dev16.dist-info → modal-1.1.4.dev17.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: modal
-Version: 1.1.4.dev16
+Version: 1.1.4.dev17
 Summary: Python client library for Modal
 Author-email: Modal Labs <support@modal.com>
 License: Apache-2.0

{modal-1.1.4.dev16.dist-info → modal-1.1.4.dev17.dist-info}/RECORD

@@ -22,7 +22,7 @@ modal/app.py,sha256=hJU3DGzP5GwYRmBj57XajljkQtLxkKftxXih2TYRcKo,48047
 modal/app.pyi,sha256=0U2xVKD3yfHe5l2bcihTDjPl__tzOvx1AIYOTebu-5o,43375
 modal/call_graph.py,sha256=1g2DGcMIJvRy-xKicuf63IVE98gJSnQsr8R_NVMptNc,2581
 modal/client.py,sha256=kyAIVB3Ay-XKJizQ_1ufUFB__EagV0MLmHJpyYyJ7J0,18636
-modal/client.pyi,sha256=oevgdxlsUFTyEK2NVmRDzB75uqZcOZrIFZEYQRJAmAg,15831
+modal/client.pyi,sha256=lxLNas06tmdopgkHnv71sy6XFQF5iYnRYkW7QoveobU,15831
 modal/cloud_bucket_mount.py,sha256=YOe9nnvSr4ZbeCn587d7_VhE9IioZYRvF9VYQTQux08,5914
 modal/cloud_bucket_mount.pyi,sha256=-qSfYAQvIoO_l2wsCCGTG5ZUwQieNKXdAO00yP1-LYU,7394
 modal/cls.py,sha256=1mBcExFrLDTZwkD3Dzu8F26_CL0CGktOV9pE60Y8g_E,40689
@@ -39,7 +39,7 @@ modal/file_io.py,sha256=OSKr77TujcXGJW1iikzYiHckLSmv07QBgBHcxxYEkoI,21456
 modal/file_io.pyi,sha256=xtO6Glf_BFwDE7QiQQo24QqcMf_Vv-iz7WojcGVlLBU,15932
 modal/file_pattern_matcher.py,sha256=A_Kdkej6q7YQyhM_2-BvpFmPqJ0oHb54B6yf9VqvPVE,8116
 modal/functions.py,sha256=kcNHvqeGBxPI7Cgd57NIBBghkfbeFJzXO44WW0jSmao,325
-modal/functions.pyi,sha256=vXmQ5-tlrxSV8E8eI0yEF2HUeF0Bd6VyQVxMCW3SpBE,38890
+modal/functions.pyi,sha256=h1oPAvbErW_i1Glzf_HsQenG-uH4KZk7X6razFbif6M,38890
 modal/gpu.py,sha256=Fe5ORvVPDIstSq1xjmM6OoNgLYFWvogP9r5BgmD3hYg,6769
 modal/image.py,sha256=nXN9k_6gApHFy8-Bk_XT2Zu3jsDsGVrC3QcuiDC4yRY,103543
 modal/image.pyi,sha256=vKdb5PpYM8wcpq9PQegeVMjrPLzAipuV4q994NZiL84,69325
@@ -150,10 +150,10 @@ modal/cli/programs/run_jupyter.py,sha256=44Lpvqk2l3hH-uOkmAOzw60NEsfB5uaRDWDKVsh
 modal/cli/programs/run_marimo.py,sha256=HyZ2za0NYqg31-mGxFQxUIAJ3Q-jRaMocEwWwI5-cdw,2887
 modal/cli/programs/vscode.py,sha256=KbTAaIXyQBVCDXxXjmBHmKpgXkUw0q4R4KkJvUjCYgk,3380
 modal/experimental/__init__.py,sha256=dPBPpxsmjZMLF3YjRrXoTvT01pl65wxi4UdFZsOem3w,15082
-modal/experimental/flash.py,sha256=7U6VuQ-7Y0HQximH4nHYq_4HDkdgRL60DE_-KRhj-Q0,19910
+modal/experimental/flash.py,sha256=gydlThBSPgS8HA53Lwi0gUAFVtpdXu20IYAQXJSoTP0,21092
 modal/experimental/flash.pyi,sha256=A8_qJGtGoXEzKDdHbvhmCw7oqfneFEvJQK3ZdTOvUdU,10830
 modal/experimental/ipython.py,sha256=TrCfmol9LGsRZMeDoeMPx3Hv3BFqQhYnmD_iH0pqdhk,2904
-modal-1.1.4.dev16.dist-info/licenses/LICENSE,sha256=psuoW8kuDP96RQsdhzwOqi6fyWv0ct8CR6Jr7He_P_k,10173
+modal-1.1.4.dev17.dist-info/licenses/LICENSE,sha256=psuoW8kuDP96RQsdhzwOqi6fyWv0ct8CR6Jr7He_P_k,10173
 modal_docs/__init__.py,sha256=svYKtV8HDwDCN86zbdWqyq5T8sMdGDj0PVlzc2tIxDM,28
 modal_docs/gen_cli_docs.py,sha256=c1yfBS_x--gL5bs0N4ihMwqwX8l3IBWSkBAKNNIi6bQ,3801
 modal_docs/gen_reference_docs.py,sha256=d_CQUGQ0rfw28u75I2mov9AlS773z9rG40-yq5o7g2U,6359
@@ -176,10 +176,10 @@ modal_proto/options_pb2.pyi,sha256=l7DBrbLO7q3Ir-XDkWsajm0d0TQqqrfuX54i4BMpdQg,1
 modal_proto/options_pb2_grpc.py,sha256=1oboBPFxaTEXt9Aw7EAj8gXHDCNMhZD2VXqocC9l_gk,159
 modal_proto/options_pb2_grpc.pyi,sha256=CImmhxHsYnF09iENPoe8S4J-n93jtgUYD2JPAc0yJSI,247
 modal_proto/py.typed,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-modal_version/__init__.py,sha256=QXpaiTjNP06yYihxFZ9XCmnyWZlsEF8DBHKKo7w2_9Y,121
+modal_version/__init__.py,sha256=N2wz9sFgRKKg4OHYFwU3AVZM3zIM6izEe5fPZoeS-LU,121
 modal_version/__main__.py,sha256=2FO0yYQQwDTh6udt1h-cBnGd1c4ZyHnHSI4BksxzVac,105
-modal-1.1.4.dev16.dist-info/METADATA,sha256=Q23DVkB7U7xPfrKgcE6akJkZ9neMGjDucWJ4j9F6-Tw,2460
-modal-1.1.4.dev16.dist-info/WHEEL,sha256=1tXe9gY0PYatrMPMDd6jXqjfpz_B-Wqm32CPfRC58XU,91
-modal-1.1.4.dev16.dist-info/entry_points.txt,sha256=An-wYgeEUnm6xzrAP9_NTSTSciYvvEWsMZILtYrvpAI,46
-modal-1.1.4.dev16.dist-info/top_level.txt,sha256=4BWzoKYREKUZ5iyPzZpjqx4G8uB5TWxXPDwibLcVa7k,43
-modal-1.1.4.dev16.dist-info/RECORD,,
+modal-1.1.4.dev17.dist-info/METADATA,sha256=otNiNfsuOjyFRzVqMRsj9VBaXbc62iXHds0Hoee0L_4,2460
+modal-1.1.4.dev17.dist-info/WHEEL,sha256=1tXe9gY0PYatrMPMDd6jXqjfpz_B-Wqm32CPfRC58XU,91
+modal-1.1.4.dev17.dist-info/entry_points.txt,sha256=An-wYgeEUnm6xzrAP9_NTSTSciYvvEWsMZILtYrvpAI,46
+modal-1.1.4.dev17.dist-info/top_level.txt,sha256=4BWzoKYREKUZ5iyPzZpjqx4G8uB5TWxXPDwibLcVa7k,43
+modal-1.1.4.dev17.dist-info/RECORD,,

modal_version/__init__.py

@@ -1,4 +1,4 @@
 # Copyright Modal Labs 2025
 """Supplies the current version of the modal client library."""
 
-__version__ = "1.1.4.dev16"
+__version__ = "1.1.4.dev17"