Use the configured region for deferred Neptune cluster tasks by SEPURI-SAI-KRISHNA · Pull Request #71646 · apache/airflow

SEPURI-SAI-KRISHNA · 2026-08-15T04:37:53Z

All three Neptune triggers declare a region_name parameter, document it, and read
self.region_name when building their hook — but never forward it to
AwsBaseWaiterTrigger.__init__, which is what actually assigns the attribute:

def__init__(self, *, db_cluster_id: str, ..., region_name: str|None=None, **kwargs):
super().__init__(
...,
aws_conn_id=aws_conn_id,
# region_name never passed**kwargs,
)
defhook(self) ->AwsGenericHook:
returnNeptuneHook(aws_conn_id=self.aws_conn_id, region_name=self.region_name, ...)

Because region_name is an explicit named parameter it is not swept up by the **kwargs
forwarding either — it is captured and discarded, so self.region_name is always None
and the deferred waiter polls the account's default region.

The failure is quiet and slow: the waiter simply never finds the cluster in the region it
is looking at, so the task burns through waiter_max_attempts and then fails with a
message that does not mention regions at all. AwsBaseWaiterTrigger.serialize() writes
self.region_name, so the wrong value also survives triggerer restarts.

There is a second, independent layer in the operators. Two of the four defer sites already
pass the hook configuration; the two in NeptuneStartDbClusterOperator.execute and
NeptuneStopDbClusterOperator.execute pass none of it, so fixing only the triggers would
still leave those paths on the default region. Both are brought in line with the existing
sites, which is why verify and botocore_config are included alongside region_name.

Affects NeptuneClusterAvailableTrigger, NeptuneClusterStoppedTrigger and
NeptuneClusterInstancesAvailableTrigger.

Tests cover both layers — that the triggers retain and serialize region_name, and that
the operators hand it to the trigger they defer with. All five fail without the change.

No newsfragment: this is a provider change, and provider changelogs are regenerated from
git log by the release manager.

Was generative AI tooling used to co-author this PR?

Yes — Claude Code (Opus 5)

Generated-by: Claude Code (Opus 5) following the guidelines

The triggers accepted a region_name but never handed it to the base trigger that assigns the attribute, and two of the operator defer sites never passed the hook configuration at all. A deferred start or stop therefore polled the default region, and the task failed only after exhausting its waiter attempts with an error that never mentioned the region.

vincbeck · 2026-08-17T21:11:21Z

@seanghaeli @ramitkataria

o-nikolas

Seems reasonable to me

A reviewer asked for the deferred trigger to be checked in full rather than on region alone, so that any future argument silently dropped on the way to the triggerer fails the suite instead of passing unnoticed.

SEPURI-SAI-KRISHNA · 2026-08-18T03:27:58Z

Thanks @ramitkataria, done in both places.

The operator test now asserts the whole serialized trigger payload instead of picking out one field:

assertdeferred.value.trigger.serialize()[1] == {
"db_cluster_id": CLUSTER_ID,
"aws_conn_id": "aws_default",
"region_name": REGION_NAME,
"verify": VERIFY,
"botocore_config": BOTOCORE_CONFIG,
"waiter_delay": WAITER_DELAY,
"waiter_max_attempts": WAITER_MAX_ATTEMPTS,
}

Exact-dict equality rather than a set of individual asserts, so an argument going missing on the way to the triggerer fails the suite, which is exactly the failure mode this PR is about. The same assertion is now in the trigger test across all three trigger classes.

One nuance the wider assertions surfaced, worth recording: of the three arguments the operators now forward, only region_name was actually being lost. verify and botocore_config are not named parameters on the Neptune trigger __init__s, so they already reached AwsBaseWaiterTrigger through **kwargs. region_nameis named, so the subclass bound it and never passed it up. The widened assertions still fail without the source change, and they now also pin down the two that were only working by accident.

Verified locally: 31 passed with the change, and exactly the 5 new parametrizations fail with it reverted (26 pre-existing tests unaffected either way).

Drafted-by: Claude Code (Opus 5); reviewed by @SEPURI-SAI-KRISHNA before posting

ramitkataria · 2026-08-18T04:30:44Z

Thanks for the addition!

SEPURI-SAI-KRISHNA · 2026-08-15T04:37:53Z

All three Neptune triggers declare a region_name parameter, document it, and read
self.region_name when building their hook — but never forward it to
AwsBaseWaiterTrigger.__init__, which is what actually assigns the attribute:

def__init__(self, *, db_cluster_id: str, ..., region_name: str|None=None, **kwargs):
super().__init__(
...,
aws_conn_id=aws_conn_id,
# region_name never passed**kwargs,
)
defhook(self) ->AwsGenericHook:
returnNeptuneHook(aws_conn_id=self.aws_conn_id, region_name=self.region_name, ...)

Because region_name is an explicit named parameter it is not swept up by the **kwargs
forwarding either — it is captured and discarded, so self.region_name is always None
and the deferred waiter polls the account's default region.

The failure is quiet and slow: the waiter simply never finds the cluster in the region it
is looking at, so the task burns through waiter_max_attempts and then fails with a
message that does not mention regions at all. AwsBaseWaiterTrigger.serialize() writes
self.region_name, so the wrong value also survives triggerer restarts.

There is a second, independent layer in the operators. Two of the four defer sites already
pass the hook configuration; the two in NeptuneStartDbClusterOperator.execute and
NeptuneStopDbClusterOperator.execute pass none of it, so fixing only the triggers would
still leave those paths on the default region. Both are brought in line with the existing
sites, which is why verify and botocore_config are included alongside region_name.

Affects NeptuneClusterAvailableTrigger, NeptuneClusterStoppedTrigger and
NeptuneClusterInstancesAvailableTrigger.

Tests cover both layers — that the triggers retain and serialize region_name, and that
the operators hand it to the trigger they defer with. All five fail without the change.

No newsfragment: this is a provider change, and provider changelogs are regenerated from
git log by the release manager.

Was generative AI tooling used to co-author this PR?

Yes — Claude Code (Opus 5)

Generated-by: Claude Code (Opus 5) following the guidelines

The triggers accepted a region_name but never handed it to the base trigger that assigns the attribute, and two of the operator defer sites never passed the hook configuration at all. A deferred start or stop therefore polled the default region, and the task failed only after exhausting its waiter attempts with an error that never mentioned the region.

vincbeck · 2026-08-17T21:11:21Z

@seanghaeli @ramitkataria

o-nikolas

Seems reasonable to me

A reviewer asked for the deferred trigger to be checked in full rather than on region alone, so that any future argument silently dropped on the way to the triggerer fails the suite instead of passing unnoticed.

SEPURI-SAI-KRISHNA · 2026-08-18T03:27:58Z

Thanks @ramitkataria, done in both places.

The operator test now asserts the whole serialized trigger payload instead of picking out one field:

assertdeferred.value.trigger.serialize()[1] == {
"db_cluster_id": CLUSTER_ID,
"aws_conn_id": "aws_default",
"region_name": REGION_NAME,
"verify": VERIFY,
"botocore_config": BOTOCORE_CONFIG,
"waiter_delay": WAITER_DELAY,
"waiter_max_attempts": WAITER_MAX_ATTEMPTS,
}

Exact-dict equality rather than a set of individual asserts, so an argument going missing on the way to the triggerer fails the suite, which is exactly the failure mode this PR is about. The same assertion is now in the trigger test across all three trigger classes.

One nuance the wider assertions surfaced, worth recording: of the three arguments the operators now forward, only region_name was actually being lost. verify and botocore_config are not named parameters on the Neptune trigger __init__s, so they already reached AwsBaseWaiterTrigger through **kwargs. region_nameis named, so the subclass bound it and never passed it up. The widened assertions still fail without the source change, and they now also pin down the two that were only working by accident.

Verified locally: 31 passed with the change, and exactly the 5 new parametrizations fail with it reverted (26 pre-existing tests unaffected either way).

Drafted-by: Claude Code (Opus 5); reviewed by @SEPURI-SAI-KRISHNA before posting

ramitkataria · 2026-08-18T04:30:44Z

Thanks for the addition!

SEPURI-SAI-KRISHNA · 2026-08-15T04:37:53Z

All three Neptune triggers declare a region_name parameter, document it, and read
self.region_name when building their hook — but never forward it to
AwsBaseWaiterTrigger.__init__, which is what actually assigns the attribute:

def__init__(self, *, db_cluster_id: str, ..., region_name: str|None=None, **kwargs):
super().__init__(
...,
aws_conn_id=aws_conn_id,
# region_name never passed**kwargs,
)
defhook(self) ->AwsGenericHook:
returnNeptuneHook(aws_conn_id=self.aws_conn_id, region_name=self.region_name, ...)

Because region_name is an explicit named parameter it is not swept up by the **kwargs
forwarding either — it is captured and discarded, so self.region_name is always None
and the deferred waiter polls the account's default region.

The failure is quiet and slow: the waiter simply never finds the cluster in the region it
is looking at, so the task burns through waiter_max_attempts and then fails with a
message that does not mention regions at all. AwsBaseWaiterTrigger.serialize() writes
self.region_name, so the wrong value also survives triggerer restarts.

There is a second, independent layer in the operators. Two of the four defer sites already
pass the hook configuration; the two in NeptuneStartDbClusterOperator.execute and
NeptuneStopDbClusterOperator.execute pass none of it, so fixing only the triggers would
still leave those paths on the default region. Both are brought in line with the existing
sites, which is why verify and botocore_config are included alongside region_name.

Affects NeptuneClusterAvailableTrigger, NeptuneClusterStoppedTrigger and
NeptuneClusterInstancesAvailableTrigger.

Tests cover both layers — that the triggers retain and serialize region_name, and that
the operators hand it to the trigger they defer with. All five fail without the change.

No newsfragment: this is a provider change, and provider changelogs are regenerated from
git log by the release manager.

Was generative AI tooling used to co-author this PR?

Yes — Claude Code (Opus 5)

Generated-by: Claude Code (Opus 5) following the guidelines

The triggers accepted a region_name but never handed it to the base trigger that assigns the attribute, and two of the operator defer sites never passed the hook configuration at all. A deferred start or stop therefore polled the default region, and the task failed only after exhausting its waiter attempts with an error that never mentioned the region.

vincbeck · 2026-08-17T21:11:21Z

@seanghaeli @ramitkataria

o-nikolas

Seems reasonable to me

A reviewer asked for the deferred trigger to be checked in full rather than on region alone, so that any future argument silently dropped on the way to the triggerer fails the suite instead of passing unnoticed.

SEPURI-SAI-KRISHNA · 2026-08-18T03:27:58Z

Thanks @ramitkataria, done in both places.

The operator test now asserts the whole serialized trigger payload instead of picking out one field:

assertdeferred.value.trigger.serialize()[1] == {
"db_cluster_id": CLUSTER_ID,
"aws_conn_id": "aws_default",
"region_name": REGION_NAME,
"verify": VERIFY,
"botocore_config": BOTOCORE_CONFIG,
"waiter_delay": WAITER_DELAY,
"waiter_max_attempts": WAITER_MAX_ATTEMPTS,
}

Exact-dict equality rather than a set of individual asserts, so an argument going missing on the way to the triggerer fails the suite, which is exactly the failure mode this PR is about. The same assertion is now in the trigger test across all three trigger classes.

One nuance the wider assertions surfaced, worth recording: of the three arguments the operators now forward, only region_name was actually being lost. verify and botocore_config are not named parameters on the Neptune trigger __init__s, so they already reached AwsBaseWaiterTrigger through **kwargs. region_nameis named, so the subclass bound it and never passed it up. The widened assertions still fail without the source change, and they now also pin down the two that were only working by accident.

Verified locally: 31 passed with the change, and exactly the 5 new parametrizations fail with it reverted (26 pre-existing tests unaffected either way).

Drafted-by: Claude Code (Opus 5); reviewed by @SEPURI-SAI-KRISHNA before posting

ramitkataria · 2026-08-18T04:30:44Z

Thanks for the addition!

SEPURI-SAI-KRISHNA · 2026-08-15T04:37:53Z

All three Neptune triggers declare a region_name parameter, document it, and read
self.region_name when building their hook — but never forward it to
AwsBaseWaiterTrigger.__init__, which is what actually assigns the attribute:

def__init__(self, *, db_cluster_id: str, ..., region_name: str|None=None, **kwargs):
super().__init__(
...,
aws_conn_id=aws_conn_id,
# region_name never passed**kwargs,
)
defhook(self) ->AwsGenericHook:
returnNeptuneHook(aws_conn_id=self.aws_conn_id, region_name=self.region_name, ...)

Because region_name is an explicit named parameter it is not swept up by the **kwargs
forwarding either — it is captured and discarded, so self.region_name is always None
and the deferred waiter polls the account's default region.

The failure is quiet and slow: the waiter simply never finds the cluster in the region it
is looking at, so the task burns through waiter_max_attempts and then fails with a
message that does not mention regions at all. AwsBaseWaiterTrigger.serialize() writes
self.region_name, so the wrong value also survives triggerer restarts.

There is a second, independent layer in the operators. Two of the four defer sites already
pass the hook configuration; the two in NeptuneStartDbClusterOperator.execute and
NeptuneStopDbClusterOperator.execute pass none of it, so fixing only the triggers would
still leave those paths on the default region. Both are brought in line with the existing
sites, which is why verify and botocore_config are included alongside region_name.

Affects NeptuneClusterAvailableTrigger, NeptuneClusterStoppedTrigger and
NeptuneClusterInstancesAvailableTrigger.

Tests cover both layers — that the triggers retain and serialize region_name, and that
the operators hand it to the trigger they defer with. All five fail without the change.

No newsfragment: this is a provider change, and provider changelogs are regenerated from
git log by the release manager.

Was generative AI tooling used to co-author this PR?

Yes — Claude Code (Opus 5)

Generated-by: Claude Code (Opus 5) following the guidelines

The triggers accepted a region_name but never handed it to the base trigger that assigns the attribute, and two of the operator defer sites never passed the hook configuration at all. A deferred start or stop therefore polled the default region, and the task failed only after exhausting its waiter attempts with an error that never mentioned the region.

vincbeck · 2026-08-17T21:11:21Z

@seanghaeli @ramitkataria

o-nikolas

Seems reasonable to me

A reviewer asked for the deferred trigger to be checked in full rather than on region alone, so that any future argument silently dropped on the way to the triggerer fails the suite instead of passing unnoticed.

SEPURI-SAI-KRISHNA · 2026-08-18T03:27:58Z

Thanks @ramitkataria, done in both places.

The operator test now asserts the whole serialized trigger payload instead of picking out one field:

assertdeferred.value.trigger.serialize()[1] == {
"db_cluster_id": CLUSTER_ID,
"aws_conn_id": "aws_default",
"region_name": REGION_NAME,
"verify": VERIFY,
"botocore_config": BOTOCORE_CONFIG,
"waiter_delay": WAITER_DELAY,
"waiter_max_attempts": WAITER_MAX_ATTEMPTS,
}

Exact-dict equality rather than a set of individual asserts, so an argument going missing on the way to the triggerer fails the suite, which is exactly the failure mode this PR is about. The same assertion is now in the trigger test across all three trigger classes.

One nuance the wider assertions surfaced, worth recording: of the three arguments the operators now forward, only region_name was actually being lost. verify and botocore_config are not named parameters on the Neptune trigger __init__s, so they already reached AwsBaseWaiterTrigger through **kwargs. region_nameis named, so the subclass bound it and never passed it up. The widened assertions still fail without the source change, and they now also pin down the two that were only working by accident.

Verified locally: 31 passed with the change, and exactly the 5 new parametrizations fail with it reverted (26 pre-existing tests unaffected either way).

Drafted-by: Claude Code (Opus 5); reviewed by @SEPURI-SAI-KRISHNA before posting

ramitkataria · 2026-08-18T04:30:44Z

Thanks for the addition!

SEPURI-SAI-KRISHNA · 2026-08-15T04:37:53Z

All three Neptune triggers declare a region_name parameter, document it, and read
self.region_name when building their hook — but never forward it to
AwsBaseWaiterTrigger.__init__, which is what actually assigns the attribute:

def__init__(self, *, db_cluster_id: str, ..., region_name: str|None=None, **kwargs):
super().__init__(
...,
aws_conn_id=aws_conn_id,
# region_name never passed**kwargs,
)
defhook(self) ->AwsGenericHook:
returnNeptuneHook(aws_conn_id=self.aws_conn_id, region_name=self.region_name, ...)

Because region_name is an explicit named parameter it is not swept up by the **kwargs
forwarding either — it is captured and discarded, so self.region_name is always None
and the deferred waiter polls the account's default region.

The failure is quiet and slow: the waiter simply never finds the cluster in the region it
is looking at, so the task burns through waiter_max_attempts and then fails with a
message that does not mention regions at all. AwsBaseWaiterTrigger.serialize() writes
self.region_name, so the wrong value also survives triggerer restarts.

There is a second, independent layer in the operators. Two of the four defer sites already
pass the hook configuration; the two in NeptuneStartDbClusterOperator.execute and
NeptuneStopDbClusterOperator.execute pass none of it, so fixing only the triggers would
still leave those paths on the default region. Both are brought in line with the existing
sites, which is why verify and botocore_config are included alongside region_name.

Affects NeptuneClusterAvailableTrigger, NeptuneClusterStoppedTrigger and
NeptuneClusterInstancesAvailableTrigger.

Tests cover both layers — that the triggers retain and serialize region_name, and that
the operators hand it to the trigger they defer with. All five fail without the change.

No newsfragment: this is a provider change, and provider changelogs are regenerated from
git log by the release manager.

Was generative AI tooling used to co-author this PR?

Yes — Claude Code (Opus 5)

Generated-by: Claude Code (Opus 5) following the guidelines

The triggers accepted a region_name but never handed it to the base trigger that assigns the attribute, and two of the operator defer sites never passed the hook configuration at all. A deferred start or stop therefore polled the default region, and the task failed only after exhausting its waiter attempts with an error that never mentioned the region.

vincbeck · 2026-08-17T21:11:21Z

@seanghaeli @ramitkataria

o-nikolas

Seems reasonable to me

A reviewer asked for the deferred trigger to be checked in full rather than on region alone, so that any future argument silently dropped on the way to the triggerer fails the suite instead of passing unnoticed.

SEPURI-SAI-KRISHNA · 2026-08-18T03:27:58Z

Thanks @ramitkataria, done in both places.

The operator test now asserts the whole serialized trigger payload instead of picking out one field:

assertdeferred.value.trigger.serialize()[1] == {
"db_cluster_id": CLUSTER_ID,
"aws_conn_id": "aws_default",
"region_name": REGION_NAME,
"verify": VERIFY,
"botocore_config": BOTOCORE_CONFIG,
"waiter_delay": WAITER_DELAY,
"waiter_max_attempts": WAITER_MAX_ATTEMPTS,
}

Exact-dict equality rather than a set of individual asserts, so an argument going missing on the way to the triggerer fails the suite, which is exactly the failure mode this PR is about. The same assertion is now in the trigger test across all three trigger classes.

One nuance the wider assertions surfaced, worth recording: of the three arguments the operators now forward, only region_name was actually being lost. verify and botocore_config are not named parameters on the Neptune trigger __init__s, so they already reached AwsBaseWaiterTrigger through **kwargs. region_nameis named, so the subclass bound it and never passed it up. The widened assertions still fail without the source change, and they now also pin down the two that were only working by accident.

Verified locally: 31 passed with the change, and exactly the 5 new parametrizations fail with it reverted (26 pre-existing tests unaffected either way).

Drafted-by: Claude Code (Opus 5); reviewed by @SEPURI-SAI-KRISHNA before posting

ramitkataria · 2026-08-18T04:30:44Z

Thanks for the addition!

SEPURI-SAI-KRISHNA · 2026-08-15T04:37:53Z

All three Neptune triggers declare a region_name parameter, document it, and read
self.region_name when building their hook — but never forward it to
AwsBaseWaiterTrigger.__init__, which is what actually assigns the attribute:

def__init__(self, *, db_cluster_id: str, ..., region_name: str|None=None, **kwargs):
super().__init__(
...,
aws_conn_id=aws_conn_id,
# region_name never passed**kwargs,
)
defhook(self) ->AwsGenericHook:
returnNeptuneHook(aws_conn_id=self.aws_conn_id, region_name=self.region_name, ...)

Because region_name is an explicit named parameter it is not swept up by the **kwargs
forwarding either — it is captured and discarded, so self.region_name is always None
and the deferred waiter polls the account's default region.

The failure is quiet and slow: the waiter simply never finds the cluster in the region it
is looking at, so the task burns through waiter_max_attempts and then fails with a
message that does not mention regions at all. AwsBaseWaiterTrigger.serialize() writes
self.region_name, so the wrong value also survives triggerer restarts.

There is a second, independent layer in the operators. Two of the four defer sites already
pass the hook configuration; the two in NeptuneStartDbClusterOperator.execute and
NeptuneStopDbClusterOperator.execute pass none of it, so fixing only the triggers would
still leave those paths on the default region. Both are brought in line with the existing
sites, which is why verify and botocore_config are included alongside region_name.

Affects NeptuneClusterAvailableTrigger, NeptuneClusterStoppedTrigger and
NeptuneClusterInstancesAvailableTrigger.

Tests cover both layers — that the triggers retain and serialize region_name, and that
the operators hand it to the trigger they defer with. All five fail without the change.

No newsfragment: this is a provider change, and provider changelogs are regenerated from
git log by the release manager.

Was generative AI tooling used to co-author this PR?

Yes — Claude Code (Opus 5)

Generated-by: Claude Code (Opus 5) following the guidelines

The triggers accepted a region_name but never handed it to the base trigger that assigns the attribute, and two of the operator defer sites never passed the hook configuration at all. A deferred start or stop therefore polled the default region, and the task failed only after exhausting its waiter attempts with an error that never mentioned the region.

vincbeck · 2026-08-17T21:11:21Z

@seanghaeli @ramitkataria

o-nikolas

Seems reasonable to me

A reviewer asked for the deferred trigger to be checked in full rather than on region alone, so that any future argument silently dropped on the way to the triggerer fails the suite instead of passing unnoticed.

SEPURI-SAI-KRISHNA · 2026-08-18T03:27:58Z

Thanks @ramitkataria, done in both places.

The operator test now asserts the whole serialized trigger payload instead of picking out one field:

assertdeferred.value.trigger.serialize()[1] == {
"db_cluster_id": CLUSTER_ID,
"aws_conn_id": "aws_default",
"region_name": REGION_NAME,
"verify": VERIFY,
"botocore_config": BOTOCORE_CONFIG,
"waiter_delay": WAITER_DELAY,
"waiter_max_attempts": WAITER_MAX_ATTEMPTS,
}

Exact-dict equality rather than a set of individual asserts, so an argument going missing on the way to the triggerer fails the suite, which is exactly the failure mode this PR is about. The same assertion is now in the trigger test across all three trigger classes.

One nuance the wider assertions surfaced, worth recording: of the three arguments the operators now forward, only region_name was actually being lost. verify and botocore_config are not named parameters on the Neptune trigger __init__s, so they already reached AwsBaseWaiterTrigger through **kwargs. region_nameis named, so the subclass bound it and never passed it up. The widened assertions still fail without the source change, and they now also pin down the two that were only working by accident.

Verified locally: 31 passed with the change, and exactly the 5 new parametrizations fail with it reverted (26 pre-existing tests unaffected either way).

Drafted-by: Claude Code (Opus 5); reviewed by @SEPURI-SAI-KRISHNA before posting

ramitkataria · 2026-08-18T04:30:44Z

Thanks for the addition!

SEPURI-SAI-KRISHNA · 2026-08-15T04:37:53Z

All three Neptune triggers declare a region_name parameter, document it, and read
self.region_name when building their hook — but never forward it to
AwsBaseWaiterTrigger.__init__, which is what actually assigns the attribute:

def__init__(self, *, db_cluster_id: str, ..., region_name: str|None=None, **kwargs):
super().__init__(
...,
aws_conn_id=aws_conn_id,
# region_name never passed**kwargs,
)
defhook(self) ->AwsGenericHook:
returnNeptuneHook(aws_conn_id=self.aws_conn_id, region_name=self.region_name, ...)

Because region_name is an explicit named parameter it is not swept up by the **kwargs
forwarding either — it is captured and discarded, so self.region_name is always None
and the deferred waiter polls the account's default region.

The failure is quiet and slow: the waiter simply never finds the cluster in the region it
is looking at, so the task burns through waiter_max_attempts and then fails with a
message that does not mention regions at all. AwsBaseWaiterTrigger.serialize() writes
self.region_name, so the wrong value also survives triggerer restarts.

There is a second, independent layer in the operators. Two of the four defer sites already
pass the hook configuration; the two in NeptuneStartDbClusterOperator.execute and
NeptuneStopDbClusterOperator.execute pass none of it, so fixing only the triggers would
still leave those paths on the default region. Both are brought in line with the existing
sites, which is why verify and botocore_config are included alongside region_name.

Affects NeptuneClusterAvailableTrigger, NeptuneClusterStoppedTrigger and
NeptuneClusterInstancesAvailableTrigger.

Tests cover both layers — that the triggers retain and serialize region_name, and that
the operators hand it to the trigger they defer with. All five fail without the change.

No newsfragment: this is a provider change, and provider changelogs are regenerated from
git log by the release manager.

Was generative AI tooling used to co-author this PR?

Yes — Claude Code (Opus 5)

Generated-by: Claude Code (Opus 5) following the guidelines

The triggers accepted a region_name but never handed it to the base trigger that assigns the attribute, and two of the operator defer sites never passed the hook configuration at all. A deferred start or stop therefore polled the default region, and the task failed only after exhausting its waiter attempts with an error that never mentioned the region.

vincbeck · 2026-08-17T21:11:21Z

@seanghaeli @ramitkataria

o-nikolas

Seems reasonable to me

A reviewer asked for the deferred trigger to be checked in full rather than on region alone, so that any future argument silently dropped on the way to the triggerer fails the suite instead of passing unnoticed.

SEPURI-SAI-KRISHNA · 2026-08-18T03:27:58Z

Thanks @ramitkataria, done in both places.

The operator test now asserts the whole serialized trigger payload instead of picking out one field:

assertdeferred.value.trigger.serialize()[1] == {
"db_cluster_id": CLUSTER_ID,
"aws_conn_id": "aws_default",
"region_name": REGION_NAME,
"verify": VERIFY,
"botocore_config": BOTOCORE_CONFIG,
"waiter_delay": WAITER_DELAY,
"waiter_max_attempts": WAITER_MAX_ATTEMPTS,
}

Exact-dict equality rather than a set of individual asserts, so an argument going missing on the way to the triggerer fails the suite, which is exactly the failure mode this PR is about. The same assertion is now in the trigger test across all three trigger classes.

One nuance the wider assertions surfaced, worth recording: of the three arguments the operators now forward, only region_name was actually being lost. verify and botocore_config are not named parameters on the Neptune trigger __init__s, so they already reached AwsBaseWaiterTrigger through **kwargs. region_nameis named, so the subclass bound it and never passed it up. The widened assertions still fail without the source change, and they now also pin down the two that were only working by accident.

Verified locally: 31 passed with the change, and exactly the 5 new parametrizations fail with it reverted (26 pre-existing tests unaffected either way).

Drafted-by: Claude Code (Opus 5); reviewed by @SEPURI-SAI-KRISHNA before posting

ramitkataria · 2026-08-18T04:30:44Z

Thanks for the addition!

SEPURI-SAI-KRISHNA · 2026-08-15T04:37:53Z

All three Neptune triggers declare a region_name parameter, document it, and read
self.region_name when building their hook — but never forward it to
AwsBaseWaiterTrigger.__init__, which is what actually assigns the attribute:

def__init__(self, *, db_cluster_id: str, ..., region_name: str|None=None, **kwargs):
super().__init__(
...,
aws_conn_id=aws_conn_id,
# region_name never passed**kwargs,
)
defhook(self) ->AwsGenericHook:
returnNeptuneHook(aws_conn_id=self.aws_conn_id, region_name=self.region_name, ...)

Because region_name is an explicit named parameter it is not swept up by the **kwargs
forwarding either — it is captured and discarded, so self.region_name is always None
and the deferred waiter polls the account's default region.

The failure is quiet and slow: the waiter simply never finds the cluster in the region it
is looking at, so the task burns through waiter_max_attempts and then fails with a
message that does not mention regions at all. AwsBaseWaiterTrigger.serialize() writes
self.region_name, so the wrong value also survives triggerer restarts.

There is a second, independent layer in the operators. Two of the four defer sites already
pass the hook configuration; the two in NeptuneStartDbClusterOperator.execute and
NeptuneStopDbClusterOperator.execute pass none of it, so fixing only the triggers would
still leave those paths on the default region. Both are brought in line with the existing
sites, which is why verify and botocore_config are included alongside region_name.

Affects NeptuneClusterAvailableTrigger, NeptuneClusterStoppedTrigger and
NeptuneClusterInstancesAvailableTrigger.

Tests cover both layers — that the triggers retain and serialize region_name, and that
the operators hand it to the trigger they defer with. All five fail without the change.

No newsfragment: this is a provider change, and provider changelogs are regenerated from
git log by the release manager.

Was generative AI tooling used to co-author this PR?

Yes — Claude Code (Opus 5)

Generated-by: Claude Code (Opus 5) following the guidelines

The triggers accepted a region_name but never handed it to the base trigger that assigns the attribute, and two of the operator defer sites never passed the hook configuration at all. A deferred start or stop therefore polled the default region, and the task failed only after exhausting its waiter attempts with an error that never mentioned the region.

vincbeck · 2026-08-17T21:11:21Z

@seanghaeli @ramitkataria

o-nikolas

Seems reasonable to me

A reviewer asked for the deferred trigger to be checked in full rather than on region alone, so that any future argument silently dropped on the way to the triggerer fails the suite instead of passing unnoticed.

SEPURI-SAI-KRISHNA · 2026-08-18T03:27:58Z

Thanks @ramitkataria, done in both places.

The operator test now asserts the whole serialized trigger payload instead of picking out one field:

assertdeferred.value.trigger.serialize()[1] == {
"db_cluster_id": CLUSTER_ID,
"aws_conn_id": "aws_default",
"region_name": REGION_NAME,
"verify": VERIFY,
"botocore_config": BOTOCORE_CONFIG,
"waiter_delay": WAITER_DELAY,
"waiter_max_attempts": WAITER_MAX_ATTEMPTS,
}

Exact-dict equality rather than a set of individual asserts, so an argument going missing on the way to the triggerer fails the suite, which is exactly the failure mode this PR is about. The same assertion is now in the trigger test across all three trigger classes.

One nuance the wider assertions surfaced, worth recording: of the three arguments the operators now forward, only region_name was actually being lost. verify and botocore_config are not named parameters on the Neptune trigger __init__s, so they already reached AwsBaseWaiterTrigger through **kwargs. region_nameis named, so the subclass bound it and never passed it up. The widened assertions still fail without the source change, and they now also pin down the two that were only working by accident.

Verified locally: 31 passed with the change, and exactly the 5 new parametrizations fail with it reverted (26 pre-existing tests unaffected either way).

Drafted-by: Claude Code (Opus 5); reviewed by @SEPURI-SAI-KRISHNA before posting

ramitkataria · 2026-08-18T04:30:44Z

Thanks for the addition!

SEPURI-SAI-KRISHNA requested a review from o-nikolas as a code owner August 15, 2026 04:37

boring-cyborgBot added area:providers provider:amazon AWS/Amazon - related issues labels Aug 15, 2026

eladkal requested a review from vincbeck August 17, 2026 20:08

o-nikolas approved these changes Aug 17, 2026
View reviewed changes

ramitkataria reviewed Aug 18, 2026
View reviewed changes

Comment threadproviders/amazon/tests/unit/amazon/aws/operators/test_neptune.py Outdated
Comment threadproviders/amazon/tests/unit/amazon/aws/triggers/test_neptune.py Outdated

Assert the whole deferred Neptune trigger payload
9a43b27
A reviewer asked for the deferred trigger to be checked in full rather than on region alone, so that any future argument silently dropped on the way to the triggerer fails the suite instead of passing unnoticed.

ramitkataria approved these changes Aug 18, 2026
View reviewed changes

o-nikolas merged commit acf0ad8 into apache:mainAug 19, 2026
83 checks passed

SEPURI-SAI-KRISHNA deleted the fix-neptune-trigger-region-name branch August 20, 2026 03:39

This was referenced Aug 20, 2026
Use the operator's AWS settings for deferred SageMaker tasks #71857
Merged
Use the operator's AWS settings for deferred Neptune, MWAA, SSM tasks #72098
Open

o-nikolas mentioned this pull request Aug 26, 2026
Status of testing Providers that were prepared on August 25, 2026 #72117
Closed
12 tasks

SEPURI-SAI-KRISHNA mentioned this pull request Aug 27, 2026
AWS deferrable operators drop hook configuration when handing off to triggers #72144
Open
2 tasks