kaxil commented on code in PR #63614: URL: https://github.com/apache/airflow/pull/63614#discussion_r3133297825
########## providers/microsoft/azure/src/airflow/providers/microsoft/azure/triggers/synapse.py: ########## @@ -0,0 +1,165 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you under the Apache License, Version 2.0 (the +# "License"); you may not use this file except in compliance +# with the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, +# software distributed under the License is distributed on an +# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +# KIND, either express or implied. See the License for the +# specific language governing permissions and limitations +# under the License. + +from __future__ import annotations + +import asyncio +import time +from collections.abc import AsyncIterator +from typing import Any + +from azure.core.exceptions import ServiceRequestError + +from airflow.providers.microsoft.azure.hooks.synapse import ( + AzureSynapsePipelineAsyncHook, + AzureSynapsePipelineRunStatus, +) +from airflow.triggers.base import BaseTrigger, TriggerEvent + + +class AzureSynapsePipelineTrigger(BaseTrigger): + """ + Trigger when an Azure Synapse pipeline run reaches a terminal state. + + If an unexpected exception occurs while polling, the trigger attempts to cancel + the pipeline run to avoid leaving orphaned executions. + + :param run_id: Pipeline run identifier. + :param azure_synapse_conn_id: Azure Synapse connection id. + :param azure_synapse_workspace_dev_endpoint: Synapse workspace dev endpoint. + :param end_time: Epoch timestamp when the trigger should timeout. + :param check_interval: Poll interval in seconds. + """ + + def __init__( + self, + run_id: str, + azure_synapse_conn_id: str, + azure_synapse_workspace_dev_endpoint: str, + end_time: float, + check_interval: int = 60, + ): + super().__init__() + + self.run_id = run_id + self.azure_synapse_conn_id = azure_synapse_conn_id + self.azure_synapse_workspace_dev_endpoint = azure_synapse_workspace_dev_endpoint + self.end_time = end_time + self.check_interval = check_interval + + def serialize(self) -> tuple[str, dict[str, Any]]: + """Serialize trigger arguments and classpath.""" + return ( + f"{self.__class__.__module__}.{self.__class__.__name__}", + { + "run_id": self.run_id, + "azure_synapse_conn_id": self.azure_synapse_conn_id, + "azure_synapse_workspace_dev_endpoint": self.azure_synapse_workspace_dev_endpoint, + "end_time": self.end_time, + "check_interval": self.check_interval, + }, + ) + + def _build_trigger_event(self, pipeline_status: str) -> TriggerEvent | None: + """ + Convert Synapse pipeline status to TriggerEvent. + + Returns None if the pipeline is still running. + """ + if pipeline_status == AzureSynapsePipelineRunStatus.SUCCEEDED: + return TriggerEvent( + { + "status": "success", + "message": f"Pipeline run {self.run_id} succeeded.", + "run_id": self.run_id, + } + ) + + if pipeline_status in AzureSynapsePipelineRunStatus.FAILURE_STATES: + return TriggerEvent( + { + "status": "failure", Review Comment: The trigger emits `{"status": "failure", ...}` for `FAILURE_STATES`, but `execute_complete` in `operators/synapse.py` only matches `"success"` and `"error"`. Any failed/cancelled pipeline lands in the `raise AzureSynapsePipelineRunException(f"Unexpected trigger event received: {event}")` fallback, so users see "Unexpected trigger event received..." instead of "Pipeline run X finished with state Failed." The sibling `AzureDataFactoryTrigger` (see `triggers/data_factory.py`) uses `"error"` for this case -- worth matching that vocabulary so the two branches of `execute_complete` cover all trigger outputs. Please add a parametrized test that round-trips a `FAILED`/`CANCELLED` trigger event through `execute_complete` and asserts the message surfaces in the raised exception. ########## providers/microsoft/azure/src/airflow/providers/microsoft/azure/triggers/synapse.py: ########## @@ -0,0 +1,165 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you under the Apache License, Version 2.0 (the +# "License"); you may not use this file except in compliance +# with the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, +# software distributed under the License is distributed on an +# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +# KIND, either express or implied. See the License for the +# specific language governing permissions and limitations +# under the License. + +from __future__ import annotations + +import asyncio +import time +from collections.abc import AsyncIterator +from typing import Any + +from azure.core.exceptions import ServiceRequestError + +from airflow.providers.microsoft.azure.hooks.synapse import ( + AzureSynapsePipelineAsyncHook, + AzureSynapsePipelineRunStatus, +) +from airflow.triggers.base import BaseTrigger, TriggerEvent + + +class AzureSynapsePipelineTrigger(BaseTrigger): + """ + Trigger when an Azure Synapse pipeline run reaches a terminal state. + + If an unexpected exception occurs while polling, the trigger attempts to cancel + the pipeline run to avoid leaving orphaned executions. + + :param run_id: Pipeline run identifier. + :param azure_synapse_conn_id: Azure Synapse connection id. + :param azure_synapse_workspace_dev_endpoint: Synapse workspace dev endpoint. + :param end_time: Epoch timestamp when the trigger should timeout. + :param check_interval: Poll interval in seconds. + """ + + def __init__( + self, + run_id: str, + azure_synapse_conn_id: str, + azure_synapse_workspace_dev_endpoint: str, + end_time: float, + check_interval: int = 60, + ): + super().__init__() + + self.run_id = run_id + self.azure_synapse_conn_id = azure_synapse_conn_id + self.azure_synapse_workspace_dev_endpoint = azure_synapse_workspace_dev_endpoint + self.end_time = end_time + self.check_interval = check_interval + + def serialize(self) -> tuple[str, dict[str, Any]]: + """Serialize trigger arguments and classpath.""" + return ( + f"{self.__class__.__module__}.{self.__class__.__name__}", + { + "run_id": self.run_id, + "azure_synapse_conn_id": self.azure_synapse_conn_id, + "azure_synapse_workspace_dev_endpoint": self.azure_synapse_workspace_dev_endpoint, + "end_time": self.end_time, + "check_interval": self.check_interval, + }, + ) + + def _build_trigger_event(self, pipeline_status: str) -> TriggerEvent | None: + """ + Convert Synapse pipeline status to TriggerEvent. + + Returns None if the pipeline is still running. + """ + if pipeline_status == AzureSynapsePipelineRunStatus.SUCCEEDED: + return TriggerEvent( + { + "status": "success", + "message": f"Pipeline run {self.run_id} succeeded.", + "run_id": self.run_id, + } + ) + + if pipeline_status in AzureSynapsePipelineRunStatus.FAILURE_STATES: + return TriggerEvent( + { + "status": "failure", + "message": f"Pipeline run {self.run_id} finished with state {pipeline_status}.", + "run_id": self.run_id, + } + ) + + return None + + async def run(self) -> AsyncIterator[TriggerEvent]: + """Poll Synapse pipeline status until completion or timeout.""" + async with AzureSynapsePipelineAsyncHook( + azure_synapse_conn_id=self.azure_synapse_conn_id, + azure_synapse_workspace_dev_endpoint=self.azure_synapse_workspace_dev_endpoint, + ) as hook: + try: + executed_after_token_refresh = True + + while time.time() < self.end_time: + try: + pipeline_status = await hook.get_pipeline_run_status(self.run_id) + + executed_after_token_refresh = True + + event = self._build_trigger_event(pipeline_status) + + if event: + yield event + return + + self.log.debug( Review Comment: Could this be `log.info` instead of `log.debug`? The ADF trigger uses info (`triggers/data_factory.py:214`), and with the triggerer's default log level users otherwise see nothing for hours while a pipeline polls. ########## providers/microsoft/azure/src/airflow/providers/microsoft/azure/operators/synapse.py: ########## @@ -255,30 +264,73 @@ def execute(self, context) -> None: context["ti"].xcom_push(key="run_id", value=self.run_id) if self.wait_for_termination: - self.log.info("Waiting for pipeline run %s to terminate.", self.run_id) - - if self.hook.wait_for_pipeline_run_status( - run_id=self.run_id, - expected_statuses=AzureSynapsePipelineRunStatus.SUCCEEDED, - check_interval=self.check_interval, - timeout=self.timeout, - ): - self.log.info("Pipeline run %s has completed successfully.", self.run_id) + if not self.deferrable: + self.log.info("Waiting for pipeline run %s to terminate.", self.run_id) + if self.hook.wait_for_pipeline_run_status( + run_id=self.run_id, + expected_statuses=AzureSynapsePipelineRunStatus.SUCCEEDED, + check_interval=self.check_interval, + timeout=self.timeout, + ): + self.log.info("Pipeline run %s has completed successfully.", self.run_id) + else: + raise AzureSynapsePipelineRunException( + f"Pipeline run {self.run_id} has failed or has been cancelled." + ) else: - raise AzureSynapsePipelineRunException( - f"Pipeline run {self.run_id} has failed or has been cancelled." + end_time = time.time() + self.timeout + + pipeline_run_status = self.hook.get_pipeline_run_status(self.run_id) + + if pipeline_run_status not in AzureSynapsePipelineRunStatus.TERMINAL_STATUSES: + self.defer( + timeout=self.execution_timeout, + trigger=AzureSynapsePipelineTrigger( + azure_synapse_conn_id=self.azure_synapse_conn_id, + azure_synapse_workspace_dev_endpoint=self.azure_synapse_workspace_dev_endpoint, + run_id=self.run_id, + check_interval=self.check_interval, + end_time=end_time, + ), + method_name=self.execute_complete.__name__, + ) + + elif pipeline_run_status == AzureSynapsePipelineRunStatus.SUCCEEDED: + self.log.info("Pipeline run %s has completed successfully.", self.run_id) + + elif pipeline_run_status in AzureSynapsePipelineRunStatus.FAILURE_STATES: + raise AzureSynapsePipelineRunException( + f"Pipeline run {self.run_id} has failed or has been cancelled." + ) + else: + if self.deferrable is True: + warnings.warn( + "Argument `wait_for_termination` is False and `deferrable` is True, " + "hence `deferrable` parameter doesn't have any effect", + UserWarning, + stacklevel=2, ) - def execute_complete(self, event: dict[str, str]) -> None: + def execute_complete(self, context: Context, event: dict[str, str] | None) -> None: """ Return immediately - callback for when the trigger fires. - Relies on trigger to throw an exception, otherwise it assumes execution was successful. + The trigger communicates the terminal pipeline state through the event payload. """ - if event: - if event["status"] == "error": - raise AirflowException(event["message"]) - self.log.info(event["message"]) + if event is None: + raise AzureSynapsePipelineRunException("Trigger returned no event.") + + status = event.get("status") + message = event.get("message", "No message returned from trigger.") + + if status == "success": + self.log.info(message) + return + + if status == "error": + raise AzureSynapsePipelineRunException(message) + + raise AzureSynapsePipelineRunException(f"Unexpected trigger event received: {event}") Review Comment: This fallback branch is reachable today because the trigger yields `status="failure"` for `FAILURE_STATES` (see `triggers/synapse.py:95`). Collapsing the trigger's `"failure"` to `"error"` would eliminate this surprise and keep the two branches above exhaustive. ########## providers/microsoft/azure/src/airflow/providers/microsoft/azure/triggers/synapse.py: ########## @@ -0,0 +1,165 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you under the Apache License, Version 2.0 (the +# "License"); you may not use this file except in compliance +# with the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, +# software distributed under the License is distributed on an +# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +# KIND, either express or implied. See the License for the +# specific language governing permissions and limitations +# under the License. + +from __future__ import annotations + +import asyncio +import time +from collections.abc import AsyncIterator +from typing import Any + +from azure.core.exceptions import ServiceRequestError + +from airflow.providers.microsoft.azure.hooks.synapse import ( + AzureSynapsePipelineAsyncHook, + AzureSynapsePipelineRunStatus, +) +from airflow.triggers.base import BaseTrigger, TriggerEvent + + +class AzureSynapsePipelineTrigger(BaseTrigger): + """ + Trigger when an Azure Synapse pipeline run reaches a terminal state. + + If an unexpected exception occurs while polling, the trigger attempts to cancel + the pipeline run to avoid leaving orphaned executions. + + :param run_id: Pipeline run identifier. + :param azure_synapse_conn_id: Azure Synapse connection id. + :param azure_synapse_workspace_dev_endpoint: Synapse workspace dev endpoint. + :param end_time: Epoch timestamp when the trigger should timeout. + :param check_interval: Poll interval in seconds. + """ + + def __init__( + self, + run_id: str, + azure_synapse_conn_id: str, + azure_synapse_workspace_dev_endpoint: str, + end_time: float, + check_interval: int = 60, + ): + super().__init__() + + self.run_id = run_id + self.azure_synapse_conn_id = azure_synapse_conn_id + self.azure_synapse_workspace_dev_endpoint = azure_synapse_workspace_dev_endpoint + self.end_time = end_time + self.check_interval = check_interval + + def serialize(self) -> tuple[str, dict[str, Any]]: + """Serialize trigger arguments and classpath.""" + return ( + f"{self.__class__.__module__}.{self.__class__.__name__}", + { + "run_id": self.run_id, + "azure_synapse_conn_id": self.azure_synapse_conn_id, + "azure_synapse_workspace_dev_endpoint": self.azure_synapse_workspace_dev_endpoint, + "end_time": self.end_time, + "check_interval": self.check_interval, + }, + ) + + def _build_trigger_event(self, pipeline_status: str) -> TriggerEvent | None: + """ + Convert Synapse pipeline status to TriggerEvent. + + Returns None if the pipeline is still running. + """ + if pipeline_status == AzureSynapsePipelineRunStatus.SUCCEEDED: + return TriggerEvent( + { + "status": "success", + "message": f"Pipeline run {self.run_id} succeeded.", + "run_id": self.run_id, + } + ) + + if pipeline_status in AzureSynapsePipelineRunStatus.FAILURE_STATES: + return TriggerEvent( + { + "status": "failure", + "message": f"Pipeline run {self.run_id} finished with state {pipeline_status}.", + "run_id": self.run_id, + } + ) + + return None + + async def run(self) -> AsyncIterator[TriggerEvent]: + """Poll Synapse pipeline status until completion or timeout.""" + async with AzureSynapsePipelineAsyncHook( + azure_synapse_conn_id=self.azure_synapse_conn_id, + azure_synapse_workspace_dev_endpoint=self.azure_synapse_workspace_dev_endpoint, + ) as hook: + try: + executed_after_token_refresh = True + + while time.time() < self.end_time: + try: + pipeline_status = await hook.get_pipeline_run_status(self.run_id) + + executed_after_token_refresh = True + + event = self._build_trigger_event(pipeline_status) + + if event: + yield event + return + + self.log.debug( + "Pipeline %s not finished yet (state=%s). Sleeping %s seconds.", + self.run_id, + pipeline_status, + self.check_interval, + ) + + await asyncio.sleep(self.check_interval) + + except ServiceRequestError: + # Azure token may expire during long pipeline runs. + if not executed_after_token_refresh: + raise + + await hook.refresh_conn() + executed_after_token_refresh = False + + yield TriggerEvent( + { + "status": "error", + "message": f"Timeout waiting for pipeline run {self.run_id}.", Review Comment: On timeout, the trigger yields an error and returns without cancelling the pipeline. The outer `except Exception` does cancel, but the normal-exit-on-deadline path doesn't reach it. Net effect: an Airflow-side timeout leaves a long-running pipeline burning Synapse compute while the task is marked failed. ADF has the same gap so this is not a regression, but consider calling `await hook.cancel_pipeline_run(self.run_id)` before the timeout yield -- or at minimum document on the operator's `timeout` parameter that the pipeline keeps running on Synapse after a trigger timeout and needs manual cancellation. ########## providers/microsoft/azure/src/airflow/providers/microsoft/azure/operators/synapse.py: ########## @@ -255,30 +264,73 @@ def execute(self, context) -> None: context["ti"].xcom_push(key="run_id", value=self.run_id) if self.wait_for_termination: - self.log.info("Waiting for pipeline run %s to terminate.", self.run_id) - - if self.hook.wait_for_pipeline_run_status( - run_id=self.run_id, - expected_statuses=AzureSynapsePipelineRunStatus.SUCCEEDED, - check_interval=self.check_interval, - timeout=self.timeout, - ): - self.log.info("Pipeline run %s has completed successfully.", self.run_id) + if not self.deferrable: + self.log.info("Waiting for pipeline run %s to terminate.", self.run_id) + if self.hook.wait_for_pipeline_run_status( + run_id=self.run_id, + expected_statuses=AzureSynapsePipelineRunStatus.SUCCEEDED, + check_interval=self.check_interval, + timeout=self.timeout, + ): + self.log.info("Pipeline run %s has completed successfully.", self.run_id) + else: + raise AzureSynapsePipelineRunException( + f"Pipeline run {self.run_id} has failed or has been cancelled." + ) else: - raise AzureSynapsePipelineRunException( - f"Pipeline run {self.run_id} has failed or has been cancelled." + end_time = time.time() + self.timeout + + pipeline_run_status = self.hook.get_pipeline_run_status(self.run_id) + + if pipeline_run_status not in AzureSynapsePipelineRunStatus.TERMINAL_STATUSES: + self.defer( + timeout=self.execution_timeout, + trigger=AzureSynapsePipelineTrigger( + azure_synapse_conn_id=self.azure_synapse_conn_id, + azure_synapse_workspace_dev_endpoint=self.azure_synapse_workspace_dev_endpoint, + run_id=self.run_id, + check_interval=self.check_interval, + end_time=end_time, + ), + method_name=self.execute_complete.__name__, + ) + + elif pipeline_run_status == AzureSynapsePipelineRunStatus.SUCCEEDED: + self.log.info("Pipeline run %s has completed successfully.", self.run_id) + + elif pipeline_run_status in AzureSynapsePipelineRunStatus.FAILURE_STATES: + raise AzureSynapsePipelineRunException( + f"Pipeline run {self.run_id} has failed or has been cancelled." + ) + else: + if self.deferrable is True: + warnings.warn( + "Argument `wait_for_termination` is False and `deferrable` is True, " + "hence `deferrable` parameter doesn't have any effect", + UserWarning, + stacklevel=2, ) - def execute_complete(self, event: dict[str, str]) -> None: + def execute_complete(self, context: Context, event: dict[str, str] | None) -> None: Review Comment: FYI this is technically a signature change for `execute_complete` (was `(self, event)`, now `(self, context, event)`). The old signature was actually wrong for the deferrable path -- `resume_execution` calls with `(context, **next_kwargs)` -- so this is a fix, not a regression. It's safe since the previous code never ran (the operator had no deferrable mode before), but worth calling out in the PR description so any downstream subclassers know. ########## providers/microsoft/azure/src/airflow/providers/microsoft/azure/operators/synapse.py: ########## @@ -255,30 +264,73 @@ def execute(self, context) -> None: context["ti"].xcom_push(key="run_id", value=self.run_id) if self.wait_for_termination: - self.log.info("Waiting for pipeline run %s to terminate.", self.run_id) - - if self.hook.wait_for_pipeline_run_status( - run_id=self.run_id, - expected_statuses=AzureSynapsePipelineRunStatus.SUCCEEDED, - check_interval=self.check_interval, - timeout=self.timeout, - ): - self.log.info("Pipeline run %s has completed successfully.", self.run_id) + if not self.deferrable: + self.log.info("Waiting for pipeline run %s to terminate.", self.run_id) + if self.hook.wait_for_pipeline_run_status( + run_id=self.run_id, + expected_statuses=AzureSynapsePipelineRunStatus.SUCCEEDED, + check_interval=self.check_interval, + timeout=self.timeout, + ): + self.log.info("Pipeline run %s has completed successfully.", self.run_id) + else: + raise AzureSynapsePipelineRunException( + f"Pipeline run {self.run_id} has failed or has been cancelled." + ) else: - raise AzureSynapsePipelineRunException( - f"Pipeline run {self.run_id} has failed or has been cancelled." + end_time = time.time() + self.timeout + + pipeline_run_status = self.hook.get_pipeline_run_status(self.run_id) + + if pipeline_run_status not in AzureSynapsePipelineRunStatus.TERMINAL_STATUSES: + self.defer( + timeout=self.execution_timeout, Review Comment: Couple of docs/packaging follow-ups for this feature that aren't in the diff yet: 1. `providers/microsoft/azure/docs/operators/azure_synapse.rst` doesn't mention the new `deferrable` parameter. `docs/operators/adf_run_pipeline.rst` has a short paragraph + exampleinclude that could be mirrored. 2. No newsfragment under `providers/microsoft/azure/newsfragments/` -- a `63614.feature.rst` noting "Add deferrable mode to `AzureSynapseRunPipelineOperator`" would be good. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
