1fanwang opened a new pull request, #72368:
URL: https://github.com/apache/airflow/pull/72368
A task retry currently submits a second Azure Synapse Spark batch even when
the first batch is still running. This can duplicate remote work.
Synchronous `AzureSynapseRunSparkBatchOperator` tasks now store the
submitted batch ID and reconnect to it on retry. A completed batch returns
immediately; a failed batch is submitted again. `wait_for_termination=False`
remains submit-only, and `durable=False` keeps the previous always-submit
behavior. Airflow versions before 3.3 continue without task-state recovery.
The shared resumable-job path also treats numeric external ID `0` as a valid
ID.
## Testing
- [x] Synapse hook and operator tests: 68 passed.
- [x] Shared `ResumableJobMixin` tests: 41 passed.
- [x] Local retry harness: the baseline submitted batches `0` and `1`; the
changed operator submitted batch `0` once, reconnected to it, and restored its
XCom.
<details>
<summary>Commands and raw output</summary>
```console
$ set -o pipefail; AIRFLOW_HOME="$(mktemp -d)/airflow" .venv/bin/uv run
--project providers/microsoft/azure pytest
providers/microsoft/azure/tests/unit/microsoft/azure/hooks/test_synapse.py
providers/microsoft/azure/tests/unit/microsoft/azure/operators/test_synapse.py
-q 2>&1 | tail -1
======================== 68 passed, 1 warning in 27.39s
========================
$ set -o pipefail; AIRFLOW_HOME="$(mktemp -d)/airflow" .venv/bin/uv run
--project task-sdk pytest
task-sdk/tests/task_sdk/bases/test_resumablejobmixin.py -q 2>&1 | tail -1
======================== 41 passed, 1 warning in 1.44s
=========================
$ AIRFLOW_HOME="$(mktemp -d)/airflow" AIRFLOW__LOGGING__LOGGING_LEVEL=ERROR
.venv/bin/uv run --project providers/microsoft/azure python - <<'PY'
import json
import subprocess
from types import ModuleType, SimpleNamespace
from unittest.mock import MagicMock
from airflow.providers.microsoft.azure.hooks.synapse import
AzureSynapseSparkBatchRunStatus
from airflow.providers.microsoft.azure.operators.synapse import
AzureSynapseRunSparkBatchOperator
PATH =
"providers/microsoft/azure/src/airflow/providers/microsoft/azure/operators/synapse.py"
class Store(dict):
def set(self, key, value):
self[key] = value
class TI:
stats_tags = {}
def __init__(self):
self.xcom = {}
def xcom_push(self, *, key, value):
self.xcom[key] = value
def load_baseline():
source = subprocess.run(
["git", "show", f"upstream/main:{PATH}"], check=True,
capture_output=True, text=True
).stdout
module = ModuleType("baseline_synapse")
module.__package__ = "airflow.providers.microsoft.azure.operators"
exec(compile(source, PATH, "exec"), module.__dict__)
return module.AzureSynapseRunSparkBatchOperator
def run(operator_type):
submissions, waited_for = [], []
store = Store()
def make_hook():
hook = MagicMock()
def submit(*, payload):
job_id = len(submissions)
submissions.append(job_id)
return SimpleNamespace(id=job_id)
hook.run_spark_job.side_effect = submit
hook.get_job_run_status.return_value =
AzureSynapseSparkBatchRunStatus.RUNNING
hook.wait_for_job_run_status.side_effect = (
lambda *, job_id, **_: waited_for.append(job_id) or True
)
return hook
retry_ti = None
retry = None
for ti in (TI(), TI()):
retry_ti = ti
retry = operator_type(
task_id=f"synapse_{len(submissions)}",
azure_synapse_conn_id="unused",
spark_pool="unused",
payload={},
check_interval=0,
)
retry.hook = make_hook()
retry.execute(context={"task_state_store": store, "ti": ti})
return {
"submissions": submissions,
"persisted_job_id": store.get("synapse_spark_batch_id"),
"retry_job_id": retry.job_id,
"retry_xcom_job_id": retry_ti.xcom["job_id"],
"waited_for": waited_for,
}
result = {"before": run(load_baseline()), "after":
run(AzureSynapseRunSparkBatchOperator)}
assert result["before"]["submissions"] == [0, 1]
assert result["after"]["submissions"] == [0]
assert result["after"]["persisted_job_id"] ==
result["after"]["retry_xcom_job_id"] == 0
print(json.dumps(result, indent=2))
PY
{
"before": {
"submissions": [0, 1],
"persisted_job_id": null,
"retry_job_id": 1,
"retry_xcom_job_id": 1,
"waited_for": [0, 1]
},
"after": {
"submissions": [0],
"persisted_job_id": 0,
"retry_job_id": 0,
"retry_xcom_job_id": 0,
"waited_for": [0, 0]
}
}
```
The retry harness used the real baseline operator, changed operator, and
`ResumableJobMixin`. It replaced Azure with an in-process hook substitute; no
Azure service was contacted.
</details>
---
##### Was generative AI tooling used to co-author this PR?
- [X] Yes - GitHub Copilot CLI (GPT-5.6 Sol)
Generated-by: GitHub Copilot CLI (GPT-5.6 Sol) following [the
guidelines](https://github.com/apache/airflow/blob/main/contributing-docs/05_pull_requests.rst#gen-ai-assisted-contributions)
---
* Read the **[Pull Request
Guidelines](https://github.com/apache/airflow/blob/main/contributing-docs/05_pull_requests.rst#pull-request-guidelines)**
for more information. Note: commit author/co-author name and email in commits
become permanently public when merged.
* For fundamental code changes, an Airflow Improvement Proposal
([AIP](https://cwiki.apache.org/confluence/display/AIRFLOW/Airflow+Improvement+Proposals))
is needed.
* When adding dependency, check compliance with the [ASF 3rd Party License
Policy](https://www.apache.org/legal/resolved.html#category-x).
* For significant user-facing changes create newsfragment:
`{pr_number}.significant.rst`, in
[airflow-core/newsfragments](https://github.com/apache/airflow/tree/main/airflow-core/newsfragments).
You can add this file in a follow-up commit after the PR is created so you
know the PR number.
--
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]
