pyworkflow-engine 0.2.13__tar.gz → 0.2.15__tar.gz

{pyworkflow_engine-0.2.13 → pyworkflow_engine-0.2.15}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: pyworkflow-engine
-Version: 0.2.13
+Version: 0.2.15
 Summary: A Python implementation of durable, event-sourced workflows inspired by Vercel Workflow
 Author: PyWorkflow Contributors
 License: MIT

{pyworkflow_engine-0.2.13 → pyworkflow_engine-0.2.15}/pyproject.toml

@@ -7,7 +7,7 @@ packages = [{include = "pyworkflow"}]
 
 [project]
 name = "pyworkflow-engine"
-version = "0.2.13"
+version = "0.2.15"
 description = "A Python implementation of durable, event-sourced workflows inspired by Vercel Workflow"
 readme = "README.md"
 requires-python = ">=3.11"

{pyworkflow_engine-0.2.13 → pyworkflow_engine-0.2.15}/pyworkflow/__init__.py

@@ -29,7 +29,7 @@ Quick Start:
     >>> run_id = await start(my_workflow, "Alice")
 """
 
-__version__ = "0.2.13"
+__version__ = "0.2.15"
 
 # Configuration
 from pyworkflow.config import (

{pyworkflow_engine-0.2.13 → pyworkflow_engine-0.2.15}/pyworkflow/celery/tasks.py

@@ -508,8 +508,17 @@ async def _record_step_completion_and_resume(
         run_id, EventType.STEP_COMPLETED.value, step_id=step_id
     )
     if already_completed:
+        # A prior worker recorded STEP_COMPLETED but may have crashed before
+        # scheduling the workflow resume. Re-schedule defensively — the resume
+        # task is idempotent via try_claim_run(SUSPENDED -> RUNNING).
+        schedule_workflow_resumption(
+            run_id,
+            datetime.now(UTC),
+            storage_config,
+            triggered_by="step_completed_idempotent_retry",
+        )
         logger.info(
-            "Step already completed by another task, skipping",
+            "Step already completed by another task, re-scheduled resume",
             run_id=run_id,
             step_id=step_id,
             step_name=step_name,
@@ -537,8 +546,16 @@ async def _record_step_completion_and_resume(
             run_id, EventType.STEP_COMPLETED.value, step_id=step_id
         )
         if already_completed:
+            # Defensive re-schedule: the peer that recorded STEP_COMPLETED may
+            # have crashed before scheduling resume. Safe due to try_claim_run.
+            schedule_workflow_resumption(
+                run_id,
+                datetime.now(UTC),
+                storage_config,
+                triggered_by="step_completed_idempotent_retry",
+            )
             logger.info(
-                "Step already completed by another task during wait, skipping",
+                "Step already completed by another task during wait, re-scheduled resume",
                 run_id=run_id,
                 step_id=step_id,
                 step_name=step_name,
@@ -563,6 +580,15 @@ async def _record_step_completion_and_resume(
     )
     await storage.record_event(completion_event)
 
+    # Release the resume_workflow_task singleton lock before scheduling the
+    # next resume. The parent resume task that dispatched this step may still
+    # be holding the (run_id,) lock during its final teardown; without this
+    # release the apply_async below sees the lock held and silently drops
+    # the schedule as a duplicate, leaving the workflow stuck in SUSPENDED.
+    resume_workflow_task.release_lock(
+        task_args=[run_id], task_kwargs={"storage_config": storage_config}
+    )
+
     # Schedule workflow resumption
     schedule_workflow_resumption(
         run_id, datetime.now(UTC), storage_config, triggered_by="step_completed"
@@ -676,8 +702,17 @@ async def _record_step_failure_and_resume(
         run_id, EventType.STEP_FAILED.value, step_id=step_id, is_retryable="False"
     )
     if already_completed or already_failed_terminal:
+        # A prior worker recorded the terminal event but may have crashed before
+        # scheduling the workflow resume. Re-schedule defensively — safe due to
+        # try_claim_run(SUSPENDED -> RUNNING) idempotency.
+        schedule_workflow_resumption(
+            run_id,
+            datetime.now(UTC),
+            storage_config,
+            triggered_by="step_failed_idempotent_retry",
+        )
         logger.info(
-            "Step already completed/failed by another task, skipping",
+            "Step already completed/failed by another task, re-scheduled resume",
             run_id=run_id,
             step_id=step_id,
             step_name=step_name,
@@ -708,8 +743,16 @@ async def _record_step_failure_and_resume(
             run_id, EventType.STEP_FAILED.value, step_id=step_id, is_retryable="False"
         )
         if already_completed or already_failed_terminal:
+            # Defensive re-schedule: peer that recorded the terminal event may
+            # have crashed before scheduling resume. Safe due to try_claim_run.
+            schedule_workflow_resumption(
+                run_id,
+                datetime.now(UTC),
+                storage_config,
+                triggered_by="step_failed_idempotent_retry",
+            )
             logger.info(
-                "Step already completed/failed by another task during wait, skipping",
+                "Step already completed/failed by another task during wait, re-scheduled resume",
                 run_id=run_id,
                 step_id=step_id,
                 step_name=step_name,
@@ -735,6 +778,12 @@ async def _record_step_failure_and_resume(
     )
     await storage.record_event(failure_event)
 
+    # Release the resume_workflow_task singleton lock before scheduling the
+    # next resume (see _record_step_completion_and_resume for details on why).
+    resume_workflow_task.release_lock(
+        task_args=[run_id], task_kwargs={"storage_config": storage_config}
+    )
+
     # Schedule workflow resumption
     schedule_workflow_resumption(
         run_id, datetime.now(UTC), storage_config, triggered_by="step_failed"

{pyworkflow_engine-0.2.13 → pyworkflow_engine-0.2.15}/pyworkflow/streams/dispatcher.py

@@ -72,9 +72,7 @@ async def dispatch_signal(signal: Signal, storage: Any) -> None:
         # running step) or by _drain_pending_signals_for_step() when the
         # step transitions back to "waiting".  Use DEBUG to avoid noisy
         # warnings for this expected scenario.
-        has_running = any(
-            s.get("status") == "running" for s in (all_subs or [])
-        )
+        has_running = any(s.get("status") == "running" for s in (all_subs or []))
         log_fn = logger.debug if has_running else logger.warning
         log_fn(
             f"No waiting steps for {signal.signal_type} on {signal.stream_id} "