always push ECS task ARN to xcom in EcsRunTaskOperator

[vandonr-amz]

that xcom value was removed in #29447 and it broke our system test that relied on that xcom value

While I agree that changing the logic around how to reattach is good, removing the xcom value is potentially a breaking change.
It used to be written only when the task didn't complete, but I think returning it all the time makes sense.

[o-nikolas]

It used to be written only when the task didn't complete, but I think returning it all the time makes sense.

We'll now have to support this new/expanded behaviour (unless/until we deprecate it of course). But on the grand scheme of things I think this is fine 👍

[Taragolis]

I have no objection for returning this value, just my 50 cents

---

This never work as it described

airflow/airflow/providers/amazon/aws/operators/ecs.py

Lines 652 to 656 in 03fcbcc

	if self.reattach:
	# Save the task ARN in XCom to be able to reattach it if needed
	self.xcom_push(context, key=self.REATTACH_XCOM_KEY, value=self.arn)

Airflow should clear all previous XComs for the task instance before execute happen

---

Initial solution might work but it save value in the "fake task_id", which might be a reason for some issues

airflow/airflow/providers/amazon/aws/operators/ecs.py

Lines 280 to 289 in fc0250f

	if self.reattach:
	# Save the task ARN in XCom to be able to reattach it if needed
	self._xcom_set(
	context,
	key=self.REATTACH_XCOM_KEY,
	value=self.arn,
	task_id=self.REATTACH_XCOM_TASK_ID_TEMPLATE.format(task_id=self.task_id),
	)

[vincbeck]

Unit tests are failing

[Taragolis]

Oh, miss that this operator never return value as it might be expected. And for work EcsRunTaskOperator -> EcsSensor you should set reattach=True and wait_for_completion = False

So make sense now

[vandonr-amz]

Oh, miss that this operator never return value as it might be expected.

yeah, the "return value" is the last line of the logs, only in some case... I don't think it's a good idea, but we have to live with it.