durable-workflow 2.0.1__tar.gz → 2.0.3__tar.gz

{durable_workflow-2.0.1/src/durable_workflow.egg-info → durable_workflow-2.0.3}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: durable-workflow
-Version: 2.0.1
+Version: 2.0.3
 Summary: Python client and worker SDK for Durable Workflow Cloud and self-hosted Server
 Author: Durable Workflow Contributors
 License-Expression: MIT

{durable_workflow-2.0.1 → durable_workflow-2.0.3}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "durable-workflow"
-version = "2.0.1"
+version = "2.0.3"
 description = "Python client and worker SDK for Durable Workflow Cloud and self-hosted Server"
 readme = "README.md"
 requires-python = ">=3.10"
@@ -71,8 +71,8 @@ durable-workflow-replay-conformance = "durable_workflow.replay_conformance:main"
 durable-workflow-workflow-updates-conformance = "durable_workflow.workflow_updates_conformance:main"
 
 [tool.durable-workflow]
-product-train = "2.0.1"
-registry-version = "2.0.1"
+product-train = "2.0.3"
+registry-version = "2.0.3"
 supported-server-versions = "2.0.0"
 worker-protocol-version = "1.19"
 control-plane-version = "2"

{durable_workflow-2.0.1 → durable_workflow-2.0.3}/src/durable_workflow/__init__.py

@@ -95,6 +95,7 @@ from .errors import (
     WorkflowNotFound,
     WorkflowPayloadDecodeError,
     WorkflowTerminated,
+    WorkflowTimedOut,
 )
 from .external_storage import (
     EXTERNAL_PAYLOAD_REFERENCE_SCHEMA,
@@ -374,6 +375,7 @@ __all__ = [
     "WorkflowFailed",
     "WorkflowNotFound",
     "WorkflowTerminated",
+    "WorkflowTimedOut",
     "EXTERNAL_TASK_INPUT_CONTRACT_SCHEMA",
     "EXTERNAL_TASK_INPUT_MEDIA_TYPE",
     "EXTERNAL_TASK_INPUT_SCHEMA",

{durable_workflow-2.0.1 → durable_workflow-2.0.3}/src/durable_workflow/client.py

@@ -46,6 +46,7 @@ from .errors import (
     WorkflowCancelled,
     WorkflowFailed,
     WorkflowTerminated,
+    WorkflowTimedOut,
     _raise_for_status,
 )
 from .external_storage import (
@@ -4141,8 +4142,9 @@ class Client:
         Raises :class:`~durable_workflow.errors.WorkflowFailed`,
         :class:`~durable_workflow.errors.WorkflowCancelled`, or
         :class:`~durable_workflow.errors.WorkflowTerminated` if the workflow
-        ended in a non-success state, or :class:`TimeoutError` if ``timeout``
-        seconds elapse before the workflow terminates.
+        ended in a non-success state. A persisted execution or run deadline
+        raises :class:`~durable_workflow.errors.WorkflowTimedOut`; a caller's
+        :class:`TimeoutError` means ``timeout`` seconds elapsed while waiting.
         """
         deadline = asyncio.get_running_loop().time() + timeout
         while True:
@@ -4177,6 +4179,8 @@ class Client:
                         raise WorkflowCancelled(
                             payload.get("reason", "workflow was cancelled")
                         )
+                    if etype == "WorkflowTimedOut":
+                        raise WorkflowTimedOut()
                 return None
             if asyncio.get_running_loop().time() > deadline:
                 raise TimeoutError(

{durable_workflow-2.0.1 → durable_workflow-2.0.3}/src/durable_workflow/errors.py

@@ -171,6 +171,32 @@ class ServerError(DurableWorkflowError):
             return self.body.get("reason")
         return None
 
+    def is_storage_admission_failure(self, poll_request_id: str | None = None) -> bool:
+        """Whether the runtime explicitly refused admission and requested an identity-preserving retry."""
+        body = self.body
+        if (
+            self.status != 503
+            or not isinstance(body, dict)
+            or self.reason() not in ("storage_pressure", "storage_admission_unavailable")
+            or body.get("retryable") is not True
+            or type(body.get("retry_after_seconds")) is not int
+            or body["retry_after_seconds"] <= 0
+            or body.get("storage_state") not in ("draining", "fenced")
+            or (self.reason() == "storage_admission_unavailable" and body["storage_state"] != "fenced")
+            or ("request_admitted" in body and body["request_admitted"] is not False)
+        ):
+            return False
+        if poll_request_id is None:
+            return body.get("request_admitted") is False
+        return (
+            bool(poll_request_id)
+            and "task" in body and body["task"] is None
+            and body.get("poll_status") == self.reason()
+            and body.get("poll_request_id") == poll_request_id
+            and body.get("retry_same_poll_request_id") is True
+            and body.get("claim_admitted") is False
+        )
+
 
 class NexusOperationFailed(DurableWorkflowError):
     """A Nexus service operation completed with a typed service failure."""
@@ -560,6 +586,17 @@ class WorkflowTerminated(DurableWorkflowError):
         super().__init__(message)
 
 
+class WorkflowTimedOut(DurableWorkflowError):
+    """A persisted workflow execution or run deadline expired.
+
+    Unlike a caller's :class:`TimeoutError` while polling, this is a terminal
+    workflow outcome recorded by the runtime.
+    """
+
+    def __init__(self, message: str = "workflow execution timed out") -> None:
+        super().__init__(message)
+
+
 class SagaCompensationFailed(DurableWorkflowError):
     """A saga compensation failed after an earlier workflow failure.
 

{durable_workflow-2.0.1 → durable_workflow-2.0.3}/src/durable_workflow/retry_policy.py

@@ -13,6 +13,9 @@
 from __future__ import annotations
 
 import asyncio
+import contextvars
+import json
+import logging
 import random
 from collections.abc import Awaitable, Callable
 from dataclasses import dataclass
@@ -20,7 +23,40 @@ from typing import TypeVar
 
 import httpx
 
+from .errors import ServerError
+
 T = TypeVar("T")
+log = logging.getLogger("durable_workflow.worker")
+
+# Task-local so sharing a Client never changes unrelated client/control requests.
+_worker_storage_admission_stop: contextvars.ContextVar[Callable[[], bool] | None] = contextvars.ContextVar(
+    "worker_storage_admission_stop", default=None,
+)
+
+
+def _storage_refusal(exc: Exception) -> tuple[ServerError, str | None] | None:
+    if not isinstance(exc, httpx.HTTPStatusError):
+        return None
+    if "X-Durable-Workflow-Protocol-Version" not in exc.request.headers:
+        return None
+    try:
+        body = exc.response.json()
+    except ValueError:
+        return None
+    error = ServerError(exc.response.status_code, body)
+    if error.reason() not in ("storage_pressure", "storage_admission_unavailable"):
+        return None
+    poll_id = None
+    if exc.request.url.path.endswith("/poll"):
+        try:
+            request = json.loads(exc.request.content)
+            poll_id = request.get("poll_request_id") if isinstance(request, dict) else None
+        except ValueError:
+            pass
+        # An invalid submitted ID must not fall through to the non-poll contract.
+        if not isinstance(poll_id, str) or not poll_id:
+            poll_id = ""
+    return error, poll_id
 
 
 @dataclass
@@ -81,6 +117,7 @@ class TransportRetryPolicy:
         Raises the last exception if all retries are exhausted.
         """
         attempt = 0
+        storage_attempt = 0
         last_exc: Exception | None = None
 
         while attempt < self.max_attempts:
@@ -89,6 +126,32 @@ class TransportRetryPolicy:
                 return result
             except Exception as exc:
                 last_exc = exc
+                stop = _worker_storage_admission_stop.get()
+                refusal = _storage_refusal(exc) if stop is not None else None
+                if refusal is not None and stop is not None:
+                    error, poll_id = refusal
+                    if not error.is_storage_admission_failure(poll_id) or stop():
+                        raise
+                    storage_attempt += 1
+                    assert isinstance(error.body, dict)
+                    delay = min(
+                        5.0,
+                        max(
+                            self.backoff_seconds(min(storage_attempt - 1, 6)),
+                            error.body["retry_after_seconds"],
+                        ),
+                    )
+                    log.warning("storage admission paused; retrying the same worker request in %.2fs", delay)
+                    # Do not consume the finite transport budget or repeat serialization/uploads.
+                    while delay > 0:
+                        if stop():
+                            raise
+                        interval = min(0.1, delay)
+                        await asyncio.sleep(interval)
+                        delay -= interval
+                    if stop():
+                        raise
+                    continue
                 if not self.should_retry(exc, attempt):
                     raise
 

{durable_workflow-2.0.1 → durable_workflow-2.0.3}/src/durable_workflow/worker.py

@@ -27,10 +27,11 @@ import time
 import traceback
 import types
 import uuid
-from collections.abc import Awaitable, Callable, Iterable, Mapping
+from collections.abc import Awaitable, Callable, Coroutine, Iterable, Mapping
 from datetime import datetime, timezone
+from functools import wraps
 from types import FunctionType
-from typing import Annotated, Any, Literal, Union, get_args, get_origin, get_type_hints
+from typing import Annotated, Any, Concatenate, Literal, ParamSpec, TypeVar, Union, get_args, get_origin, get_type_hints
 
 from . import serializer
 from .activity import ActivityContext, ActivityInfo, _set_context
@@ -75,6 +76,7 @@ from .metrics import (
     WORKER_TASKS,
     MetricsRecorder,
 )
+from .retry_policy import _worker_storage_admission_stop
 from .workflow import (
     Command,
     NexusServiceCall,
@@ -137,6 +139,27 @@ _WORKFLOW_TASK_COMPLETION_MAX_ATTEMPTS = 3
 _WORKFLOW_TASK_COMPLETION_RETRY_DELAYS = (0.05, 0.2)
 _WORKFLOW_TASK_NEXUS_RESOLUTION_LIMIT = 100
 _WORKER_WORKFLOW_FINGERPRINTS: dict[tuple[str, str], str] = {}
+_P = ParamSpec("_P")
+_R = TypeVar("_R")
+
+
+def _with_storage_admission_retries(
+    fn: Callable[Concatenate[Worker, _P], Coroutine[Any, Any, _R]],
+) -> Callable[Concatenate[Worker, _P], Coroutine[Any, Any, _R]]:
+    @wraps(fn)
+    async def run(self: Worker, /, *args: _P.args, **kwargs: _P.kwargs) -> _R:
+        token = _worker_storage_admission_stop.set(self._stop.is_set)
+        try:
+            return await fn(self, *args, **kwargs)
+        finally:
+            _worker_storage_admission_stop.reset(token)
+    return run
+
+
+def _is_storage_admission_error(error: BaseException) -> bool:
+    return isinstance(error, ServerError) and error.reason() in (
+        "storage_pressure", "storage_admission_unavailable",
+    )
 
 
 def _command_payload_codec(codec: object) -> str:
@@ -208,6 +231,8 @@ def _should_fail_workflow_task_after_completion_error(error: BaseException) -> b
 
 
 def _should_retry_workflow_task_completion_error(error: BaseException) -> bool:
+    if _is_storage_admission_error(error):
+        return False
     if isinstance(error, ServerError):
         return error.status >= 500 or error.status == 429
 
@@ -1424,6 +1449,8 @@ class Worker:
                 )
             except Exception as e:
                 log.warning("failed to complete workflow update task %s: %s", task_id, e)
+                if _is_storage_admission_error(e):
+                    return None
                 if _should_fail_workflow_task_after_completion_error(e):
                     await self._report_workflow_task_after_completion_error(task_id, attempt, e)
                     return None
@@ -1550,6 +1577,8 @@ class Worker:
             )
         except Exception as e:
             log.warning("failed to complete workflow task %s: %s", task_id, e)
+            if _is_storage_admission_error(e):
+                return None
             if _should_fail_workflow_task_after_completion_error(e):
                 await self._report_workflow_task_after_completion_error(task_id, attempt, e)
                 return None
@@ -1736,6 +1765,8 @@ class Worker:
                 log.warning("failed to report activity failure: %s", fe)
             return "failed_non_retryable"
         except Exception as e:
+            if _is_storage_admission_error(e):
+                raise
             log.exception("activity failed")
             try:
                 await self.client.fail_activity_task(
@@ -1971,6 +2002,9 @@ class Worker:
                 **self._external_storage_completion_kwargs(),
             )
         except ServerError as e:
+            if _is_storage_admission_error(e):
+                log.warning("query task %s acknowledgement paused: %s", query_task_id, e)
+                return "complete_error"
             if _is_final_query_task_rejection(e):
                 log.info(
                     "query task %s completion was rejected after the task ended server-side: %s",
@@ -2147,6 +2181,8 @@ class Worker:
                 self._release_workflow_capacity()
                 if self._stop.is_set():
                     return
+                if _is_storage_admission_error(e):
+                    raise
                 self._record_poll_metrics("workflow", "error", time.perf_counter() - poll_start)
                 log.warning("workflow poll error: %s", e)
                 await asyncio.sleep(1.0)
@@ -2236,10 +2272,15 @@ class Worker:
                     timeout=self._poll_http_timeout,
                     build_id=self.build_id,
                 )
+            except asyncio.CancelledError:
+                self._act_semaphore.release()
+                raise
             except Exception as e:
                 self._act_semaphore.release()
                 if self._stop.is_set():
                     return
+                if _is_storage_admission_error(e):
+                    raise
                 self._record_poll_metrics("activity", "error", time.perf_counter() - poll_start)
                 log.warning("activity poll error: %s", e)
                 await asyncio.sleep(1.0)
@@ -2289,6 +2330,8 @@ class Worker:
                     query_thread_stop is not None and query_thread_stop.is_set()
                 ):
                     return
+                if _is_storage_admission_error(e):
+                    raise
                 self._record_poll_metrics("query", "error", time.perf_counter() - poll_start)
                 log.warning("query poll error: %s", e)
                 await asyncio.sleep(1.0)
@@ -2525,6 +2568,7 @@ class Worker:
         except Exception:
             log.exception("query task poller thread stopped unexpectedly")
 
+    @_with_storage_admission_retries
     async def _query_task_thread_main(self) -> None:
         loop = asyncio.get_running_loop()
         task = asyncio.current_task()
@@ -2562,6 +2606,7 @@ class Worker:
                 "the worker registration remains active"
             )
 
+    @_with_storage_admission_retries
     async def run(self) -> None:
         """Register the worker and poll until `stop()` is called or the task is cancelled."""
         self._begin_run()
@@ -2622,6 +2667,8 @@ class Worker:
                     process_metrics=self._current_process_metrics(),
                 )
             except Exception as e:
+                if _is_storage_admission_error(e) and not self._stop.is_set():
+                    raise
                 log.warning("worker heartbeat failed: %s", e)
                 continue
             if isinstance(ack, dict):
@@ -2703,6 +2750,7 @@ class Worker:
 
         return metrics
 
+    @_with_storage_admission_retries
     async def run_until(
         self,
         *,
@@ -2714,17 +2762,19 @@ class Worker:
 
         This is intended for examples, smoke tests, and single-workflow scripts.
         Long-running workers should call :meth:`run` and coordinate shutdown from
-        their process supervisor.
+        their process supervisor. ``timeout`` includes registration and runtime
+        admission pauses; shutdown retains its separate drain timeout.
         """
         background_tasks: list[asyncio.Task[Any]] = []
+        deadline = asyncio.get_running_loop().time() + timeout
 
         self._begin_run()
         try:
-            await self._register()
-        finally:
-            self._registration_done.set()
+            try:
+                await asyncio.wait_for(self._register(), timeout=timeout)
+            finally:
+                self._registration_done.set()
 
-        try:
             if self._stop.is_set():
                 raise asyncio.CancelledError
             background_tasks.append(asyncio.create_task(self._heartbeat_loop()))
@@ -2742,7 +2792,11 @@ class Worker:
                 )
             )
             self._poller_tasks.add(run_until_loop)
-            return await run_until_loop
+            return await asyncio.wait_for(
+                run_until_loop, timeout=max(0.0, deadline - asyncio.get_running_loop().time()),
+            )
+        except asyncio.TimeoutError as error:
+            raise TimeoutError(f"workflow {workflow_id} not terminal after {timeout}s") from error
         finally:
             primary_error = sys.exc_info()[1]
             primary_traceback = primary_error.__traceback__ if primary_error is not None else None

{durable_workflow-2.0.1 → durable_workflow-2.0.3/src/durable_workflow.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: durable-workflow
-Version: 2.0.1
+Version: 2.0.3
 Summary: Python client and worker SDK for Durable Workflow Cloud and self-hosted Server
 Author: Durable Workflow Contributors
 License-Expression: MIT

{durable_workflow-2.0.1 → durable_workflow-2.0.3}/src/durable_workflow.egg-info/SOURCES.txt

@@ -73,11 +73,13 @@ tests/test_serializer.py
 tests/test_signals.py
 tests/test_sleep.py
 tests/test_standalone_activity_client.py
+tests/test_storage_admission.py
 tests/test_sync.py
 tests/test_testing_harness.py
 tests/test_update_signal_condition_replay.py
 tests/test_updates.py
 tests/test_wait_condition.py
 tests/test_worker.py
+tests/test_workflow_result_timeout.py
 tests/test_workflow_streams.py
 tests/test_workflow_updates_conformance.py

{durable_workflow-2.0.1 → durable_workflow-2.0.3}/tests/test_pypi_project_surface.py

@@ -348,7 +348,7 @@ def test_stable_audit_requires_root_and_bare_install(
     assert evidence["historical_versions"] == ["0.4.105", "0.4.106"]
 
 
-def test_stable_root_preserves_non_yanked_historical_files() -> None:
+def test_stable_root_preserves_historical_files_without_overriding_yanks() -> None:
     source = source_metadata(stable=True)
     exact_urls = verify_exact_version_json(exact_version_payload(source), source)
 
@@ -356,8 +356,7 @@ def test_stable_root_preserves_non_yanked_historical_files() -> None:
 
     payload = stable_project_payload()
     payload["releases"]["0.4.106"][0]["yanked"] = True
-    with pytest.raises(ProjectSurfaceError, match="historical release 0.4.106 must remain retained and non-yanked"):
-        verify_stable_project_json(payload, source, exact_urls)
+    assert verify_stable_project_json(payload, source, exact_urls) == ("0.4.105", "0.4.106")
 
 
 def test_prerelease_evidence_records_deferred_default_surface(

{durable_workflow-2.0.1 → durable_workflow-2.0.3}/tests/test_release_metadata.py

@@ -41,9 +41,9 @@ def test_worker_release_identity_matches_supported_server_and_protocol() -> None
     project = manifest["project"]
     release = manifest["tool"]["durable-workflow"]
 
-    assert project["version"] == "2.0.1"
+    assert project["version"] == "2.0.3"
     assert release["product-train"] == project["version"]
-    assert release["registry-version"] == "2.0.1"
+    assert release["registry-version"] == "2.0.3"
     assert release["supported-server-versions"] == "2.0.0"
     assert release["worker-protocol-version"] == PROTOCOL_VERSION == "1.19"
     assert release["durable-selection"] is True