Preserve event resource version after empty sync polls by hkc-8010 · Pull Request #66471 · apache/airflow

hkc-8010 · 2026-05-06T12:20:45Z

What this PR does

This fixes the synchronous KubernetesPodOperator event polling path when the first
list_namespaced_event response is empty.

Today PodManager.watch_pod_events() only advances resource_version from individual
event items. If the first response contains no items, the loop keeps calling
list_namespaced_event without a resource_version, which can lead to repeated
unversioned event LISTs during pod startup.

This change seeds resource_version from the list response metadata immediately after
read_pod_events() returns, so subsequent polls can use
resource_version_match="NotOlderThan" even when the first response is empty.

Why this is still needed

Recent kubernetes provider fixes improved event handling in related areas, but they do
not cover this specific synchronous empty-first-response path in PodManager.

Tests

Added a regression test covering:

first event list response has no items
response metadata includes resource_version
second poll uses that resource_version
second poll sets resource_version_match="NotOlderThan"

Local validation

Ran locally:

prek run --from-ref upstream/main --to-ref HEAD
breeze testing providers-tests --backend postgres --python 3.10 --db-reset providers/cncf/kubernetes/tests/unit/cncf/kubernetes/utils/test_pod_manager.py providers/cncf/kubernetes/tests/unit/cncf/kubernetes/hooks/test_kubernetes.py
breeze testing providers-tests --backend sqlite --python 3.10 --db-reset --force-lowest-dependencies providers/cncf/kubernetes/tests/unit/cncf/kubernetes/utils/test_pod_manager.py providers/cncf/kubernetes/tests/unit/cncf/kubernetes/hooks/test_kubernetes.py

Note: pushing from the fork required bypassing a pre-push hook failure in dev/registry
that also reproduces on clean upstream/main and is unrelated to this PR.

Gen-AI disclosure

This PR was created with assistance from generative AI tools. I reviewed and validated
the generated changes locally, including static checks and targeted Breeze test runs.

hkc-8010 · 2026-05-06T15:09:22Z

I checked the four failing provider distributions tests / Compat ... jobs, and they all fail on the same unrelated test in the grouped amazon provider batch:

providers/amazon/tests/unit/amazon/aws/hooks/test_sagemaker.py::TestSageMakerHook::test_delete_model_when_not_exist
E assert 400 == 404

I also verified the same failure is already present on a recent scheduled main run:

run 25429216179
job provider distributions tests / Compat 3.1.8:P3.10: (74592770298)
same assert 400 == 404 failure in test_sagemaker.py

So these compat failures look inherited from main, not introduced by this cncf.kubernetes change.

jscheffl · 2026-05-06T19:07:24Z

I checked the four failing provider distributions tests / Compat ... jobs, and they all fail on the same unrelated test in the grouped amazon provider batch:
providers/amazon/tests/unit/amazon/aws/hooks/test_sagemaker.py::TestSageMakerHook::test_delete_model_when_not_existE assert 400 == 404
I also verified the same failure is already present on a recent scheduled main run:
* run `25429216179`
* job `provider distributions tests / Compat 3.1.8:P3.10:` (`74592770298`)
* same `assert 400 == 404` failure in `test_sagemaker.py`
So these compat failures look inherited from main, not introduced by this cncf.kubernetes change.

Was just fixed in #66473 - rebasing...

… empty sync polls Closesapache#66470

jscheffl · 2026-05-06T20:19:05Z

Note: As we do not use semantic commits changed title.

boring-cyborg · 2026-05-06T20:19:29Z

Awesome work, congrats on your first merged pull request! You are invited to check our Issue Tracker for additional contributions.

hkc-8010 · 2026-05-06T12:20:45Z

Closes#66470

What this PR does

This fixes the synchronous KubernetesPodOperator event polling path when the first
list_namespaced_event response is empty.

Today PodManager.watch_pod_events() only advances resource_version from individual
event items. If the first response contains no items, the loop keeps calling
list_namespaced_event without a resource_version, which can lead to repeated
unversioned event LISTs during pod startup.

This change seeds resource_version from the list response metadata immediately after
read_pod_events() returns, so subsequent polls can use
resource_version_match="NotOlderThan" even when the first response is empty.

Why this is still needed

Recent kubernetes provider fixes improved event handling in related areas, but they do
not cover this specific synchronous empty-first-response path in PodManager.

Tests

Added a regression test covering:

first event list response has no items
response metadata includes resource_version
second poll uses that resource_version
second poll sets resource_version_match="NotOlderThan"

Local validation

Ran locally:

prek run --from-ref upstream/main --to-ref HEAD
breeze testing providers-tests --backend postgres --python 3.10 --db-reset providers/cncf/kubernetes/tests/unit/cncf/kubernetes/utils/test_pod_manager.py providers/cncf/kubernetes/tests/unit/cncf/kubernetes/hooks/test_kubernetes.py
breeze testing providers-tests --backend sqlite --python 3.10 --db-reset --force-lowest-dependencies providers/cncf/kubernetes/tests/unit/cncf/kubernetes/utils/test_pod_manager.py providers/cncf/kubernetes/tests/unit/cncf/kubernetes/hooks/test_kubernetes.py

Note: pushing from the fork required bypassing a pre-push hook failure in dev/registry
that also reproduces on clean upstream/main and is unrelated to this PR.

Gen-AI disclosure

This PR was created with assistance from generative AI tools. I reviewed and validated
the generated changes locally, including static checks and targeted Breeze test runs.

hkc-8010 · 2026-05-06T15:09:22Z

I checked the four failing provider distributions tests / Compat ... jobs, and they all fail on the same unrelated test in the grouped amazon provider batch:

providers/amazon/tests/unit/amazon/aws/hooks/test_sagemaker.py::TestSageMakerHook::test_delete_model_when_not_exist
E assert 400 == 404

I also verified the same failure is already present on a recent scheduled main run:

run 25429216179
job provider distributions tests / Compat 3.1.8:P3.10: (74592770298)
same assert 400 == 404 failure in test_sagemaker.py

So these compat failures look inherited from main, not introduced by this cncf.kubernetes change.

jscheffl · 2026-05-06T19:07:24Z

I checked the four failing provider distributions tests / Compat ... jobs, and they all fail on the same unrelated test in the grouped amazon provider batch:
providers/amazon/tests/unit/amazon/aws/hooks/test_sagemaker.py::TestSageMakerHook::test_delete_model_when_not_existE assert 400 == 404
I also verified the same failure is already present on a recent scheduled main run:
* run `25429216179`
* job `provider distributions tests / Compat 3.1.8:P3.10:` (`74592770298`)
* same `assert 400 == 404` failure in `test_sagemaker.py`
So these compat failures look inherited from main, not introduced by this cncf.kubernetes change.

Was just fixed in #66473 - rebasing...

… empty sync polls Closesapache#66470

jscheffl · 2026-05-06T20:19:05Z

Note: As we do not use semantic commits changed title.

boring-cyborg · 2026-05-06T20:19:29Z

Awesome work, congrats on your first merged pull request! You are invited to check our Issue Tracker for additional contributions.

hkc-8010 · 2026-05-06T12:20:45Z

Closes#66470

What this PR does

This fixes the synchronous KubernetesPodOperator event polling path when the first
list_namespaced_event response is empty.

Today PodManager.watch_pod_events() only advances resource_version from individual
event items. If the first response contains no items, the loop keeps calling
list_namespaced_event without a resource_version, which can lead to repeated
unversioned event LISTs during pod startup.

This change seeds resource_version from the list response metadata immediately after
read_pod_events() returns, so subsequent polls can use
resource_version_match="NotOlderThan" even when the first response is empty.

Why this is still needed

Recent kubernetes provider fixes improved event handling in related areas, but they do
not cover this specific synchronous empty-first-response path in PodManager.

Tests

Added a regression test covering:

first event list response has no items
response metadata includes resource_version
second poll uses that resource_version
second poll sets resource_version_match="NotOlderThan"

Local validation

Ran locally:

prek run --from-ref upstream/main --to-ref HEAD
breeze testing providers-tests --backend postgres --python 3.10 --db-reset providers/cncf/kubernetes/tests/unit/cncf/kubernetes/utils/test_pod_manager.py providers/cncf/kubernetes/tests/unit/cncf/kubernetes/hooks/test_kubernetes.py
breeze testing providers-tests --backend sqlite --python 3.10 --db-reset --force-lowest-dependencies providers/cncf/kubernetes/tests/unit/cncf/kubernetes/utils/test_pod_manager.py providers/cncf/kubernetes/tests/unit/cncf/kubernetes/hooks/test_kubernetes.py

Note: pushing from the fork required bypassing a pre-push hook failure in dev/registry
that also reproduces on clean upstream/main and is unrelated to this PR.

Gen-AI disclosure

This PR was created with assistance from generative AI tools. I reviewed and validated
the generated changes locally, including static checks and targeted Breeze test runs.

hkc-8010 · 2026-05-06T15:09:22Z

I checked the four failing provider distributions tests / Compat ... jobs, and they all fail on the same unrelated test in the grouped amazon provider batch:

providers/amazon/tests/unit/amazon/aws/hooks/test_sagemaker.py::TestSageMakerHook::test_delete_model_when_not_exist
E assert 400 == 404

I also verified the same failure is already present on a recent scheduled main run:

run 25429216179
job provider distributions tests / Compat 3.1.8:P3.10: (74592770298)
same assert 400 == 404 failure in test_sagemaker.py

So these compat failures look inherited from main, not introduced by this cncf.kubernetes change.

jscheffl · 2026-05-06T19:07:24Z

I checked the four failing provider distributions tests / Compat ... jobs, and they all fail on the same unrelated test in the grouped amazon provider batch:
providers/amazon/tests/unit/amazon/aws/hooks/test_sagemaker.py::TestSageMakerHook::test_delete_model_when_not_existE assert 400 == 404
I also verified the same failure is already present on a recent scheduled main run:
* run `25429216179`
* job `provider distributions tests / Compat 3.1.8:P3.10:` (`74592770298`)
* same `assert 400 == 404` failure in `test_sagemaker.py`
So these compat failures look inherited from main, not introduced by this cncf.kubernetes change.

Was just fixed in #66473 - rebasing...

… empty sync polls Closesapache#66470

jscheffl · 2026-05-06T20:19:05Z

Note: As we do not use semantic commits changed title.

boring-cyborg · 2026-05-06T20:19:29Z

Awesome work, congrats on your first merged pull request! You are invited to check our Issue Tracker for additional contributions.

hkc-8010 · 2026-05-06T12:20:45Z

Closes#66470

What this PR does

This fixes the synchronous KubernetesPodOperator event polling path when the first
list_namespaced_event response is empty.

Today PodManager.watch_pod_events() only advances resource_version from individual
event items. If the first response contains no items, the loop keeps calling
list_namespaced_event without a resource_version, which can lead to repeated
unversioned event LISTs during pod startup.

This change seeds resource_version from the list response metadata immediately after
read_pod_events() returns, so subsequent polls can use
resource_version_match="NotOlderThan" even when the first response is empty.

Why this is still needed

Recent kubernetes provider fixes improved event handling in related areas, but they do
not cover this specific synchronous empty-first-response path in PodManager.

Tests

Added a regression test covering:

first event list response has no items
response metadata includes resource_version
second poll uses that resource_version
second poll sets resource_version_match="NotOlderThan"

Local validation

Ran locally:

prek run --from-ref upstream/main --to-ref HEAD
breeze testing providers-tests --backend postgres --python 3.10 --db-reset providers/cncf/kubernetes/tests/unit/cncf/kubernetes/utils/test_pod_manager.py providers/cncf/kubernetes/tests/unit/cncf/kubernetes/hooks/test_kubernetes.py
breeze testing providers-tests --backend sqlite --python 3.10 --db-reset --force-lowest-dependencies providers/cncf/kubernetes/tests/unit/cncf/kubernetes/utils/test_pod_manager.py providers/cncf/kubernetes/tests/unit/cncf/kubernetes/hooks/test_kubernetes.py

Note: pushing from the fork required bypassing a pre-push hook failure in dev/registry
that also reproduces on clean upstream/main and is unrelated to this PR.

Gen-AI disclosure

This PR was created with assistance from generative AI tools. I reviewed and validated
the generated changes locally, including static checks and targeted Breeze test runs.

hkc-8010 · 2026-05-06T15:09:22Z

I checked the four failing provider distributions tests / Compat ... jobs, and they all fail on the same unrelated test in the grouped amazon provider batch:

providers/amazon/tests/unit/amazon/aws/hooks/test_sagemaker.py::TestSageMakerHook::test_delete_model_when_not_exist
E assert 400 == 404

I also verified the same failure is already present on a recent scheduled main run:

run 25429216179
job provider distributions tests / Compat 3.1.8:P3.10: (74592770298)
same assert 400 == 404 failure in test_sagemaker.py

So these compat failures look inherited from main, not introduced by this cncf.kubernetes change.

jscheffl · 2026-05-06T19:07:24Z

I checked the four failing provider distributions tests / Compat ... jobs, and they all fail on the same unrelated test in the grouped amazon provider batch:
providers/amazon/tests/unit/amazon/aws/hooks/test_sagemaker.py::TestSageMakerHook::test_delete_model_when_not_existE assert 400 == 404
I also verified the same failure is already present on a recent scheduled main run:
* run `25429216179`
* job `provider distributions tests / Compat 3.1.8:P3.10:` (`74592770298`)
* same `assert 400 == 404` failure in `test_sagemaker.py`
So these compat failures look inherited from main, not introduced by this cncf.kubernetes change.

Was just fixed in #66473 - rebasing...

… empty sync polls Closesapache#66470

jscheffl · 2026-05-06T20:19:05Z

Note: As we do not use semantic commits changed title.

boring-cyborg · 2026-05-06T20:19:29Z

Awesome work, congrats on your first merged pull request! You are invited to check our Issue Tracker for additional contributions.

hkc-8010 · 2026-05-06T12:20:45Z

Closes#66470

What this PR does

This fixes the synchronous KubernetesPodOperator event polling path when the first
list_namespaced_event response is empty.

Today PodManager.watch_pod_events() only advances resource_version from individual
event items. If the first response contains no items, the loop keeps calling
list_namespaced_event without a resource_version, which can lead to repeated
unversioned event LISTs during pod startup.

This change seeds resource_version from the list response metadata immediately after
read_pod_events() returns, so subsequent polls can use
resource_version_match="NotOlderThan" even when the first response is empty.

Why this is still needed

Recent kubernetes provider fixes improved event handling in related areas, but they do
not cover this specific synchronous empty-first-response path in PodManager.

Tests

Added a regression test covering:

first event list response has no items
response metadata includes resource_version
second poll uses that resource_version
second poll sets resource_version_match="NotOlderThan"

Local validation

Ran locally:

prek run --from-ref upstream/main --to-ref HEAD
breeze testing providers-tests --backend postgres --python 3.10 --db-reset providers/cncf/kubernetes/tests/unit/cncf/kubernetes/utils/test_pod_manager.py providers/cncf/kubernetes/tests/unit/cncf/kubernetes/hooks/test_kubernetes.py
breeze testing providers-tests --backend sqlite --python 3.10 --db-reset --force-lowest-dependencies providers/cncf/kubernetes/tests/unit/cncf/kubernetes/utils/test_pod_manager.py providers/cncf/kubernetes/tests/unit/cncf/kubernetes/hooks/test_kubernetes.py

Note: pushing from the fork required bypassing a pre-push hook failure in dev/registry
that also reproduces on clean upstream/main and is unrelated to this PR.

Gen-AI disclosure

This PR was created with assistance from generative AI tools. I reviewed and validated
the generated changes locally, including static checks and targeted Breeze test runs.

hkc-8010 · 2026-05-06T15:09:22Z

I checked the four failing provider distributions tests / Compat ... jobs, and they all fail on the same unrelated test in the grouped amazon provider batch:

providers/amazon/tests/unit/amazon/aws/hooks/test_sagemaker.py::TestSageMakerHook::test_delete_model_when_not_exist
E assert 400 == 404

I also verified the same failure is already present on a recent scheduled main run:

run 25429216179
job provider distributions tests / Compat 3.1.8:P3.10: (74592770298)
same assert 400 == 404 failure in test_sagemaker.py

So these compat failures look inherited from main, not introduced by this cncf.kubernetes change.

jscheffl · 2026-05-06T19:07:24Z

I checked the four failing provider distributions tests / Compat ... jobs, and they all fail on the same unrelated test in the grouped amazon provider batch:
providers/amazon/tests/unit/amazon/aws/hooks/test_sagemaker.py::TestSageMakerHook::test_delete_model_when_not_existE assert 400 == 404
I also verified the same failure is already present on a recent scheduled main run:
* run `25429216179`
* job `provider distributions tests / Compat 3.1.8:P3.10:` (`74592770298`)
* same `assert 400 == 404` failure in `test_sagemaker.py`
So these compat failures look inherited from main, not introduced by this cncf.kubernetes change.

Was just fixed in #66473 - rebasing...

… empty sync polls Closesapache#66470

jscheffl · 2026-05-06T20:19:05Z

Note: As we do not use semantic commits changed title.

boring-cyborg · 2026-05-06T20:19:29Z

Awesome work, congrats on your first merged pull request! You are invited to check our Issue Tracker for additional contributions.

hkc-8010 · 2026-05-06T12:20:45Z

Closes#66470

What this PR does

This fixes the synchronous KubernetesPodOperator event polling path when the first
list_namespaced_event response is empty.

Today PodManager.watch_pod_events() only advances resource_version from individual
event items. If the first response contains no items, the loop keeps calling
list_namespaced_event without a resource_version, which can lead to repeated
unversioned event LISTs during pod startup.

This change seeds resource_version from the list response metadata immediately after
read_pod_events() returns, so subsequent polls can use
resource_version_match="NotOlderThan" even when the first response is empty.

Why this is still needed

Recent kubernetes provider fixes improved event handling in related areas, but they do
not cover this specific synchronous empty-first-response path in PodManager.

Tests

Added a regression test covering:

first event list response has no items
response metadata includes resource_version
second poll uses that resource_version
second poll sets resource_version_match="NotOlderThan"

Local validation

Ran locally:

prek run --from-ref upstream/main --to-ref HEAD
breeze testing providers-tests --backend postgres --python 3.10 --db-reset providers/cncf/kubernetes/tests/unit/cncf/kubernetes/utils/test_pod_manager.py providers/cncf/kubernetes/tests/unit/cncf/kubernetes/hooks/test_kubernetes.py
breeze testing providers-tests --backend sqlite --python 3.10 --db-reset --force-lowest-dependencies providers/cncf/kubernetes/tests/unit/cncf/kubernetes/utils/test_pod_manager.py providers/cncf/kubernetes/tests/unit/cncf/kubernetes/hooks/test_kubernetes.py

Note: pushing from the fork required bypassing a pre-push hook failure in dev/registry
that also reproduces on clean upstream/main and is unrelated to this PR.

Gen-AI disclosure

This PR was created with assistance from generative AI tools. I reviewed and validated
the generated changes locally, including static checks and targeted Breeze test runs.

hkc-8010 · 2026-05-06T15:09:22Z

I checked the four failing provider distributions tests / Compat ... jobs, and they all fail on the same unrelated test in the grouped amazon provider batch:

providers/amazon/tests/unit/amazon/aws/hooks/test_sagemaker.py::TestSageMakerHook::test_delete_model_when_not_exist
E assert 400 == 404

I also verified the same failure is already present on a recent scheduled main run:

run 25429216179
job provider distributions tests / Compat 3.1.8:P3.10: (74592770298)
same assert 400 == 404 failure in test_sagemaker.py

So these compat failures look inherited from main, not introduced by this cncf.kubernetes change.

jscheffl · 2026-05-06T19:07:24Z

I checked the four failing provider distributions tests / Compat ... jobs, and they all fail on the same unrelated test in the grouped amazon provider batch:
providers/amazon/tests/unit/amazon/aws/hooks/test_sagemaker.py::TestSageMakerHook::test_delete_model_when_not_existE assert 400 == 404
I also verified the same failure is already present on a recent scheduled main run:
* run `25429216179`
* job `provider distributions tests / Compat 3.1.8:P3.10:` (`74592770298`)
* same `assert 400 == 404` failure in `test_sagemaker.py`
So these compat failures look inherited from main, not introduced by this cncf.kubernetes change.

Was just fixed in #66473 - rebasing...

… empty sync polls Closesapache#66470

jscheffl · 2026-05-06T20:19:05Z

Note: As we do not use semantic commits changed title.

boring-cyborg · 2026-05-06T20:19:29Z

Awesome work, congrats on your first merged pull request! You are invited to check our Issue Tracker for additional contributions.

hkc-8010 · 2026-05-06T12:20:45Z

Closes#66470

What this PR does

This fixes the synchronous KubernetesPodOperator event polling path when the first
list_namespaced_event response is empty.

Today PodManager.watch_pod_events() only advances resource_version from individual
event items. If the first response contains no items, the loop keeps calling
list_namespaced_event without a resource_version, which can lead to repeated
unversioned event LISTs during pod startup.

This change seeds resource_version from the list response metadata immediately after
read_pod_events() returns, so subsequent polls can use
resource_version_match="NotOlderThan" even when the first response is empty.

Why this is still needed

Recent kubernetes provider fixes improved event handling in related areas, but they do
not cover this specific synchronous empty-first-response path in PodManager.

Tests

Added a regression test covering:

first event list response has no items
response metadata includes resource_version
second poll uses that resource_version
second poll sets resource_version_match="NotOlderThan"

Local validation

Ran locally:

prek run --from-ref upstream/main --to-ref HEAD
breeze testing providers-tests --backend postgres --python 3.10 --db-reset providers/cncf/kubernetes/tests/unit/cncf/kubernetes/utils/test_pod_manager.py providers/cncf/kubernetes/tests/unit/cncf/kubernetes/hooks/test_kubernetes.py
breeze testing providers-tests --backend sqlite --python 3.10 --db-reset --force-lowest-dependencies providers/cncf/kubernetes/tests/unit/cncf/kubernetes/utils/test_pod_manager.py providers/cncf/kubernetes/tests/unit/cncf/kubernetes/hooks/test_kubernetes.py

Note: pushing from the fork required bypassing a pre-push hook failure in dev/registry
that also reproduces on clean upstream/main and is unrelated to this PR.

Gen-AI disclosure

This PR was created with assistance from generative AI tools. I reviewed and validated
the generated changes locally, including static checks and targeted Breeze test runs.

hkc-8010 · 2026-05-06T15:09:22Z

I checked the four failing provider distributions tests / Compat ... jobs, and they all fail on the same unrelated test in the grouped amazon provider batch:

providers/amazon/tests/unit/amazon/aws/hooks/test_sagemaker.py::TestSageMakerHook::test_delete_model_when_not_exist
E assert 400 == 404

I also verified the same failure is already present on a recent scheduled main run:

run 25429216179
job provider distributions tests / Compat 3.1.8:P3.10: (74592770298)
same assert 400 == 404 failure in test_sagemaker.py

So these compat failures look inherited from main, not introduced by this cncf.kubernetes change.

jscheffl · 2026-05-06T19:07:24Z

I checked the four failing provider distributions tests / Compat ... jobs, and they all fail on the same unrelated test in the grouped amazon provider batch:
providers/amazon/tests/unit/amazon/aws/hooks/test_sagemaker.py::TestSageMakerHook::test_delete_model_when_not_existE assert 400 == 404
I also verified the same failure is already present on a recent scheduled main run:
* run `25429216179`
* job `provider distributions tests / Compat 3.1.8:P3.10:` (`74592770298`)
* same `assert 400 == 404` failure in `test_sagemaker.py`
So these compat failures look inherited from main, not introduced by this cncf.kubernetes change.

Was just fixed in #66473 - rebasing...

… empty sync polls Closesapache#66470

jscheffl · 2026-05-06T20:19:05Z

Note: As we do not use semantic commits changed title.

boring-cyborg · 2026-05-06T20:19:29Z

Awesome work, congrats on your first merged pull request! You are invited to check our Issue Tracker for additional contributions.

hkc-8010 · 2026-05-06T12:20:45Z

Closes#66470

What this PR does

This fixes the synchronous KubernetesPodOperator event polling path when the first
list_namespaced_event response is empty.

Today PodManager.watch_pod_events() only advances resource_version from individual
event items. If the first response contains no items, the loop keeps calling
list_namespaced_event without a resource_version, which can lead to repeated
unversioned event LISTs during pod startup.

This change seeds resource_version from the list response metadata immediately after
read_pod_events() returns, so subsequent polls can use
resource_version_match="NotOlderThan" even when the first response is empty.

Why this is still needed

Recent kubernetes provider fixes improved event handling in related areas, but they do
not cover this specific synchronous empty-first-response path in PodManager.

Tests

Added a regression test covering:

first event list response has no items
response metadata includes resource_version
second poll uses that resource_version
second poll sets resource_version_match="NotOlderThan"

Local validation

Ran locally:

prek run --from-ref upstream/main --to-ref HEAD
breeze testing providers-tests --backend postgres --python 3.10 --db-reset providers/cncf/kubernetes/tests/unit/cncf/kubernetes/utils/test_pod_manager.py providers/cncf/kubernetes/tests/unit/cncf/kubernetes/hooks/test_kubernetes.py
breeze testing providers-tests --backend sqlite --python 3.10 --db-reset --force-lowest-dependencies providers/cncf/kubernetes/tests/unit/cncf/kubernetes/utils/test_pod_manager.py providers/cncf/kubernetes/tests/unit/cncf/kubernetes/hooks/test_kubernetes.py

Note: pushing from the fork required bypassing a pre-push hook failure in dev/registry
that also reproduces on clean upstream/main and is unrelated to this PR.

Gen-AI disclosure

This PR was created with assistance from generative AI tools. I reviewed and validated
the generated changes locally, including static checks and targeted Breeze test runs.

hkc-8010 · 2026-05-06T15:09:22Z

I checked the four failing provider distributions tests / Compat ... jobs, and they all fail on the same unrelated test in the grouped amazon provider batch:

providers/amazon/tests/unit/amazon/aws/hooks/test_sagemaker.py::TestSageMakerHook::test_delete_model_when_not_exist
E assert 400 == 404

I also verified the same failure is already present on a recent scheduled main run:

run 25429216179
job provider distributions tests / Compat 3.1.8:P3.10: (74592770298)
same assert 400 == 404 failure in test_sagemaker.py

So these compat failures look inherited from main, not introduced by this cncf.kubernetes change.

jscheffl · 2026-05-06T19:07:24Z

I checked the four failing provider distributions tests / Compat ... jobs, and they all fail on the same unrelated test in the grouped amazon provider batch:
providers/amazon/tests/unit/amazon/aws/hooks/test_sagemaker.py::TestSageMakerHook::test_delete_model_when_not_existE assert 400 == 404
I also verified the same failure is already present on a recent scheduled main run:
* run `25429216179`
* job `provider distributions tests / Compat 3.1.8:P3.10:` (`74592770298`)
* same `assert 400 == 404` failure in `test_sagemaker.py`
So these compat failures look inherited from main, not introduced by this cncf.kubernetes change.

Was just fixed in #66473 - rebasing...

… empty sync polls Closesapache#66470

jscheffl · 2026-05-06T20:19:05Z

Note: As we do not use semantic commits changed title.

boring-cyborg · 2026-05-06T20:19:29Z

Awesome work, congrats on your first merged pull request! You are invited to check our Issue Tracker for additional contributions.

boring-cyborgBot added area:providers provider:cncf-kubernetes Kubernetes (k8s) provider related issues labels May 6, 2026

hkc-8010 marked this pull request as ready for review May 6, 2026 12:32

hkc-8010 requested review from hussein-awala, jedcunningham and jscheffl as code owners May 6, 2026 12:32

fix(providers-cncf-kubernetes): preserve event resource version after…
da82928
… empty sync polls Closesapache#66470

jscheffl force-pushed the fix/kpo-sync-event-resource-version branch from 93632ad to da82928Compare May 6, 2026 19:07

jscheffl changed the title ~~fix(providers-cncf-kubernetes): preserve event resource version after empty sync polls~~Preserve event resource version after empty sync pollsMay 6, 2026

jscheffl approved these changes May 6, 2026
View reviewed changes

jscheffl merged commit 7ca4c75 into apache:mainMay 6, 2026
113 checks passed

jscheffl mentioned this pull request May 19, 2026
Status of testing Providers that were prepared on May 19, 2026 #67213
Closed