edda-framework 0.13.0__py3-none-any.whl → 0.14.1__py3-none-any.whl

edda/app.py

@@ -583,7 +583,6 @@ class EddaApp:
                 auto_resume_stale_workflows_periodically(
                     self.storage,
                     self.replay_engine,
-                    self.worker_id,
                     interval=60,
                 ),
                 name="leader_stale_workflow_resume",
@@ -628,7 +627,6 @@ class EddaApp:
                 auto_resume_stale_workflows_periodically(
                     self.storage,
                     self.replay_engine,
-                    self.worker_id,
                     interval=60,
                 ),
                 name="leader_stale_workflow_resume",
@@ -1411,7 +1409,8 @@ class EddaApp:
         from growing indefinitely with orphaned messages (messages that were
         published but never received by any subscriber).
 
-        Uses system-level locking to ensure only one pod executes cleanup at a time.
+        Important: This task should only be run by a single worker (e.g., via leader
+        election). It does not perform its own distributed coordination.
 
         Args:
             interval: Cleanup interval in seconds (default: 3600 = 1 hour)
@@ -1422,27 +1421,13 @@ class EddaApp:
         """
         while True:
             try:
-                # Add jitter to prevent thundering herd in multi-pod deployments
+                # Add jitter to prevent thundering herd
                 jitter = random.uniform(0, interval * 0.3)
                 await asyncio.sleep(interval + jitter)
 
-                # Try to acquire global lock for this task
-                lock_acquired = await self.storage.try_acquire_system_lock(
-                    lock_name="cleanup_old_messages",
-                    worker_id=self.worker_id,
-                    timeout_seconds=interval,
-                )
-
-                if not lock_acquired:
-                    # Another pod is handling this task
-                    continue
-
-                try:
-                    deleted_count = await self.storage.cleanup_old_channel_messages(retention_days)
-                    if deleted_count > 0:
-                        logger.info("Cleaned up %d old channel messages", deleted_count)
-                finally:
-                    await self.storage.release_system_lock("cleanup_old_messages", self.worker_id)
+                deleted_count = await self.storage.cleanup_old_channel_messages(retention_days)
+                if deleted_count > 0:
+                    logger.info("Cleaned up %d old channel messages", deleted_count)
             except Exception as e:
                 logger.error("Error cleaning up old messages: %s", e, exc_info=True)
 

edda/channels.py

@@ -129,6 +129,24 @@ class WaitForTimerException(Exception):
         super().__init__(f"Waiting for timer: {timer_id}")
 
 
+class ChannelModeConflictError(Exception):
+    """
+    Raised when subscribing with a different mode than the channel's established mode.
+
+    A channel's mode is locked when the first subscription is created. Subsequent
+    subscriptions must use the same mode.
+    """
+
+    def __init__(self, channel: str, existing_mode: str, requested_mode: str) -> None:
+        self.channel = channel
+        self.existing_mode = existing_mode
+        self.requested_mode = requested_mode
+        super().__init__(
+            f"Channel '{channel}' is already configured as '{existing_mode}' mode. "
+            f"Cannot subscribe with '{requested_mode}' mode."
+        )
+
+
 # =============================================================================
 # Subscription Functions
 # =============================================================================
@@ -150,6 +168,10 @@ async def subscribe(
               - "competing": Each message goes to only one subscriber (work queue pattern)
               - "direct": Receive messages sent via send_to() to this instance
 
+    Raises:
+        ChannelModeConflictError: If the channel is already configured with a different mode
+        ValueError: If mode is not 'broadcast', 'competing', or 'direct'
+
     The "direct" mode is syntactic sugar that subscribes to "channel:instance_id" internally,
     allowing simpler code when receiving direct messages:
 
@@ -204,6 +226,11 @@ async def subscribe(
             f"Invalid subscription mode: {mode}. Must be 'broadcast', 'competing', or 'direct'"
         )
 
+    # Check for mode conflict
+    existing_mode = await ctx.storage.get_channel_mode(actual_channel)
+    if existing_mode is not None and existing_mode != actual_mode:
+        raise ChannelModeConflictError(channel, existing_mode, mode)
+
     await ctx.storage.subscribe_to_channel(ctx.instance_id, actual_channel, actual_mode)
 
 

edda/locking.py

@@ -192,7 +192,6 @@ async def _refresh_lock_periodically(
 
 async def cleanup_stale_locks_periodically(
     storage: StorageProtocol,
-    worker_id: str,
     interval: int = 60,
 ) -> None:
     """
@@ -204,49 +203,37 @@ async def cleanup_stale_locks_periodically(
     Note: This function only cleans up locks without resuming workflows.
     For automatic workflow resumption, use auto_resume_stale_workflows_periodically().
 
-    Uses system-level locking to ensure only one pod executes cleanup at a time.
+    Important: This function should only be run by a single worker (e.g., via leader
+    election). It does not perform its own distributed coordination.
 
     Example:
         >>> asyncio.create_task(
-        ...     cleanup_stale_locks_periodically(storage, worker_id, interval=60)
+        ...     cleanup_stale_locks_periodically(storage, interval=60)
         ... )
 
     Args:
         storage: Storage backend
-        worker_id: Unique identifier for this worker (for global lock coordination)
         interval: Cleanup interval in seconds (default: 60)
     """
     with suppress(asyncio.CancelledError):
         while True:
-            # Add jitter to prevent thundering herd in multi-pod deployments
+            # Add jitter to prevent thundering herd
             jitter = random.uniform(0, interval * 0.3)
             await asyncio.sleep(interval + jitter)
 
-            # Try to acquire global lock for this task
-            lock_acquired = await storage.try_acquire_system_lock(
-                lock_name="cleanup_stale_locks",
-                worker_id=worker_id,
-                timeout_seconds=interval,
-            )
-
-            if not lock_acquired:
-                # Another pod is handling this task
-                continue
-
             try:
                 # Clean up stale locks
                 workflows = await storage.cleanup_stale_locks()
 
                 if len(workflows) > 0:
                     logger.info("Cleaned up %d stale locks", len(workflows))
-            finally:
-                await storage.release_system_lock("cleanup_stale_locks", worker_id)
+            except Exception as e:
+                logger.error("Failed to cleanup stale locks: %s", e, exc_info=True)
 
 
 async def auto_resume_stale_workflows_periodically(
     storage: StorageProtocol,
     replay_engine: Any,
-    worker_id: str,
     interval: int = 60,
 ) -> None:
     """
@@ -255,39 +242,27 @@ async def auto_resume_stale_workflows_periodically(
     This combines lock cleanup with automatic workflow resumption, ensuring
     that workflows interrupted by worker crashes are automatically recovered.
 
-    Uses system-level locking to ensure only one pod executes this task at a time,
-    preventing duplicate workflow execution (CRITICAL for safety).
+    Important: This function should only be run by a single worker (e.g., via leader
+    election). It does not perform its own distributed coordination.
 
     Example:
         >>> asyncio.create_task(
         ...     auto_resume_stale_workflows_periodically(
-        ...         storage, replay_engine, worker_id, interval=60
+        ...         storage, replay_engine, interval=60
         ...     )
         ... )
 
     Args:
         storage: Storage backend
         replay_engine: ReplayEngine instance for resuming workflows
-        worker_id: Unique identifier for this worker (for global lock coordination)
         interval: Cleanup interval in seconds (default: 60)
     """
     with suppress(asyncio.CancelledError):
         while True:
-            # Add jitter to prevent thundering herd in multi-pod deployments
+            # Add jitter to prevent thundering herd
             jitter = random.uniform(0, interval * 0.3)
             await asyncio.sleep(interval + jitter)
 
-            # Try to acquire global lock for this task
-            lock_acquired = await storage.try_acquire_system_lock(
-                lock_name="auto_resume_stale_workflows",
-                worker_id=worker_id,
-                timeout_seconds=interval,
-            )
-
-            if not lock_acquired:
-                # Another pod is handling this task
-                continue
-
             try:
                 # Clean up stale locks and get workflows to resume
                 workflows_to_resume = await storage.cleanup_stale_locks()
@@ -369,8 +344,8 @@ async def auto_resume_stale_workflows_periodically(
                                 e,
                                 exc_info=True,
                             )
-            finally:
-                await storage.release_system_lock("auto_resume_stale_workflows", worker_id)
+            except Exception as e:
+                logger.error("Failed to cleanup stale locks: %s", e, exc_info=True)
 
 
 class LockNotAcquiredError(Exception):

edda/storage/protocol.py

@@ -990,6 +990,18 @@ class StorageProtocol(Protocol):
         """
         ...
 
+    async def get_channel_mode(self, channel: str) -> str | None:
+        """
+        Get the mode for a channel (from any existing subscription).
+
+        Args:
+            channel: Channel name
+
+        Returns:
+            The mode ('broadcast' or 'competing') or None if no subscriptions exist
+        """
+        ...
+
     async def register_channel_receive_and_release_lock(
         self,
         instance_id: str,

edda/storage/sqlalchemy_storage.py

@@ -3170,6 +3170,26 @@ class SQLAlchemyStorage:
                 "cursor_message_id": subscription.cursor_message_id,
             }
 
+    async def get_channel_mode(self, channel: str) -> str | None:
+        """
+        Get the mode for a channel (from any existing subscription).
+
+        Args:
+            channel: Channel name
+
+        Returns:
+            The mode ('broadcast' or 'competing') or None if no subscriptions exist
+        """
+        session = self._get_session_for_operation()
+        async with self._session_scope(session) as session:
+            result = await session.execute(
+                select(ChannelSubscription.mode)
+                .where(ChannelSubscription.channel == channel)
+                .limit(1)
+            )
+            row = result.scalar_one_or_none()
+            return row
+
     async def register_channel_receive_and_release_lock(
         self,
         instance_id: str,

{edda_framework-0.13.0.dist-info → edda_framework-0.14.1.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: edda-framework
-Version: 0.13.0
+Version: 0.14.1
 Summary: Lightweight Durable Execution Framework
 Project-URL: Homepage, https://github.com/i2y/edda
 Project-URL: Documentation, https://github.com/i2y/edda#readme

{edda_framework-0.13.0.dist-info → edda_framework-0.14.1.dist-info}/RECORD

@@ -1,12 +1,12 @@
 edda/__init__.py,sha256=hGC6WR2R36M8LWC97F-0Rw4Ln0QUUT_1xC-7acOy_Fk,2237
 edda/activity.py,sha256=nRm9eBrr0lFe4ZRQ2whyZ6mo5xd171ITIVhqytUhOpw,21025
-edda/app.py,sha256=ITTc7x5S4ykCP3KPZXKxuNczXkPtbn04ZQaxcem46Hw,68406
-edda/channels.py,sha256=CosFoB9HVHBKRmhU_t6qoCV3l6egAGt3sqpakfgZLKc,36596
+edda/app.py,sha256=gtBtNsWpib8BHwzC02MlP7LxUHNXqcZVwtFcUhTWGkk,67801
+edda/channels.py,sha256=6JFZkeOs0xDumexr0_bLI_Mb4S245hLJM_Sqp3xPCCA,37676
 edda/compensation.py,sha256=iKLlnTxiF1YSatmYQW84EkPB1yMKUEZBtgjuGnghLtY,11824
 edda/context.py,sha256=Qqm_nUC5NNnOfHAb7taqKqZVIc0GoRWUrjZ4L9_-q70,22128
 edda/exceptions.py,sha256=-ntBLGpVQgPFG5N1o8m_7weejAYkNrUdxTkOP38vsHk,1766
 edda/hooks.py,sha256=HUZ6FTM__DZjwuomDfTDEroQ3mugEPuJHcGm7CTQNvg,8193
-edda/locking.py,sha256=NAFJmw-JaSVsXn4Y4czJyv_s9bWG8cdrzDBWIEag5X8,13661
+edda/locking.py,sha256=ZMdzGO4u3h8m3kDysmkDpAkJNvQQADbILPsmo52EQis,12716
 edda/pydantic_utils.py,sha256=dGVPNrrttDeq1k233PopCtjORYjZitsgASPfPnO6R10,9056
 edda/replay.py,sha256=IQGByw9mlTpRulyUgsHJSPsZUULmM2YqFcm2WeB4jtw,43227
 edda/retry.py,sha256=t4_E1skrhotA1XWHTLbKi-DOgCMasOUnhI9OT-O_eCE,6843
@@ -34,8 +34,8 @@ edda/storage/migrations.py,sha256=KrceouVODct9WWDBhmjAW0IYptDWd2mqJmhrHnee59M,13
 edda/storage/models.py,sha256=axXGJ-Orwcd_AsEUwIyFfDyg3NQxMcOQ2mrTzXkNv3g,12284
 edda/storage/notify_base.py,sha256=gUb-ypG1Bo0c-KrleYmC7eKtdwQNUeqGS5k7UILlSsQ,5055
 edda/storage/pg_notify.py,sha256=myzJ9xX86uiro9aaiA1SW1sN3E-zYafn7_lpeAy1jOg,11830
-edda/storage/protocol.py,sha256=vdB5GvBen8lgUA0qEfBXfQTLbVfGKeBTQuEwSUqLZtI,39463
-edda/storage/sqlalchemy_storage.py,sha256=HREK7fHmq3DGx6w4jA03_NrQu9HbyMomyIawMuOQLYQ,146246
+edda/storage/protocol.py,sha256=tLUbD7SQ71oJVaTKfeh5HG1hvuLfaxoqC-a8m-iF0LY,39786
+edda/storage/sqlalchemy_storage.py,sha256=UQmq3C_iC2j3N7q2V0kPcbtdwnFHAtyYWd9NBHVWsWQ,146934
 edda/viewer_ui/__init__.py,sha256=N1-T33SXadOXcBsDSgJJ9Iqz4y4verJngWryQu70c5c,517
 edda/viewer_ui/app.py,sha256=xZdIIGX5D2efNWQSVpPdldxLukHHpJD7JiAa_YKG5Uw,97084
 edda/viewer_ui/components.py,sha256=A0IxLwgj_Lu51O57OfzOwME8jzoJtKegEVvSnWc7uPo,45174
@@ -47,8 +47,8 @@ edda/visualizer/mermaid_generator.py,sha256=XWa2egoOTNDfJEjPcwoxwQmblUqXf7YInWFj
 edda/migrations/mysql/20251217000000_initial_schema.sql,sha256=LpINasESRhadOeqABwDk4JZ0OZ4_zQw_opnhIR4Xe9U,12367
 edda/migrations/postgresql/20251217000000_initial_schema.sql,sha256=hCaGMWeptpzpnsjfNKVsMYuwPRe__fK9E0VZpClAumQ,11732
 edda/migrations/sqlite/20251217000000_initial_schema.sql,sha256=Wq9gCnQ0K9SOt0PY_8f1MG4va8rLVWIIcf2lnRzSK5g,11906
-edda_framework-0.13.0.dist-info/METADATA,sha256=K-ar-0liixJ38d34eDvPGYPyWxUSKjgsQ_xwZPCbZ2A,37567
-edda_framework-0.13.0.dist-info/WHEEL,sha256=WLgqFyCfm_KASv4WHyYy0P3pM_m7J5L9k2skdKLirC8,87
-edda_framework-0.13.0.dist-info/entry_points.txt,sha256=dPH47s6UoJgUZxHoeSMqZsQkLaSE-SGLi-gh88k2WrU,48
-edda_framework-0.13.0.dist-info/licenses/LICENSE,sha256=udxb-V7_cYKTHqW7lNm48rxJ-Zpf0WAY_PyGDK9BPCo,1069
-edda_framework-0.13.0.dist-info/RECORD,,
+edda_framework-0.14.1.dist-info/METADATA,sha256=3WamC1lB2LrLdUIbOPrYeoWsqW8leTXF8zRFB8rObpY,37567
+edda_framework-0.14.1.dist-info/WHEEL,sha256=WLgqFyCfm_KASv4WHyYy0P3pM_m7J5L9k2skdKLirC8,87
+edda_framework-0.14.1.dist-info/entry_points.txt,sha256=dPH47s6UoJgUZxHoeSMqZsQkLaSE-SGLi-gh88k2WrU,48
+edda_framework-0.14.1.dist-info/licenses/LICENSE,sha256=udxb-V7_cYKTHqW7lNm48rxJ-Zpf0WAY_PyGDK9BPCo,1069
+edda_framework-0.14.1.dist-info/RECORD,,