Status of testing Providers that were prepared on October 02, 2026

[shahar1]

I have a kind request for all the contributors to the latest provider distributions release.
Could you please help us to test the RC versions of the providers?

The guidelines on how to test providers can be found in

Verify providers by contributors

Let us know in the comments, whether the issue is addressed.

These are providers that require testing as there were some substantial changes introduced:

Provider cncf.kubernetes: 10.23.0rc2

• Template every connection id accepted by provider operators (#73286): @shahar1
  Linked issues:
  • Linked Issue #70296: @shahar1
  • Linked Issue #35259: @kimminw00
• Fix KubernetesExecutor not queuing tasks on Airflow 3.0.x (#73764): @anishgirianish
  Linked issues:
  • Linked Issue #73658: @shahar1
  • Linked Issue #63491: @anishgirianish
• Prevent KubernetesExecutor from launching stale workloads (#69762): @wolvery
  Linked issues:
  • Linked Issue #69760: @wolvery
• Fix KubernetesPodOperator discarding successful XCom when sidecar kill fails (#72068): @shoemoney
  Linked issues:
  • Linked Issue #71369: @vinay-kumar-uppala
• Fix deferrable Kubernetes 401s with a default exec-based kubeconfig (#72300): @rjgoyln
  Linked issues:
  • Linked Issue #61737: @andrewhharmon
  • Linked Issue #65212: @amoghrajesh
• Fix unformatted Kubernetes pod trigger error messages (#73034): @rjgoyln
  Linked issues:
  • Linked Issue #72697: @rjgoyln
• Apply KubernetesPodOperator template fields only after rendering (#71174): @rjgoyln
  Linked issues:
  • Linked Issue #70296: @shahar1
• Unify executor workload queues (#63491): @anishgirianish
• Remove mention of CeleryKubernetesExecutor from KubernetesExecutor doc (#73624): @alant94
• Document that clear_on_success also removes the KPO pod identity (#71749): @thc1006
  Linked issues:
  • Linked Issue #71743: @thc1006
• Remove note about airflow 2.7.0 from docs (#72913): @eladkal

Provider duckdb: 0.2.0rc1 🎉 New provider

Note: This is a new provider, so the PRs below are collected from the git history of the provider's sources rather than from a changelog diff. The list may therefore include bootstrap and development commits (e.g. release-preparation commits) that are not user-facing changes.

• Release the DuckDB provider as 0.2.0 (#73888): @shahar1
• Add DuckDB support to the Amazon provider (#73301): @o-nikolas
• Add DuckDB provider (#72845): @o-nikolas

Provider edge3: 5.0.0rc2

• Make EdgeExecutor respect [core] parallelism (#72048): @FrankYang0529
  Linked issues:
  • Linked Issue #51009: @kaxil
• Add OIDC worker token verification to Edge3 (#72262): @robertpofuk
• Fix scheduler crash on restart with queued or running Edge tasks (#73832): @FrankYang0529
  Linked issues:
  • Linked Issue #72048: @FrankYang0529
• Fix airflow edge list-workers always showing null concurrency (#72959): @Eason09053360
  Linked issues:
  • Linked Issue #65472: @jscheffl
• Unify executor workload queues (#63491): @anishgirianish
• Render the database ERD as a searchable Mermaid diagram instead of an image (#72006): @ColtenOuO
  Linked issues:
  • Linked Issue #36842: @BasPH
• Clarify that AccessView.JOBS is the Edge worker management permission (#72627): @potiuk
  Linked issues:
  • Linked Issue #67379: @jscheffl

All users involved in the PRs:
@thc1006 @eladkal @o-nikolas @wolvery @robertpofuk @shoemoney @shahar1 @Eason09053360 @rjgoyln @ColtenOuO @anishgirianish @alant94 @potiuk @FrankYang0529

[FrankYang0529]

Tested edge3 5.0.0rc2 and it works as expected:

• Fix scheduler crash on restart with queued or running Edge tasks #73832: Restarted the scheduler on Airflow 3.3.2 while Edge tasks were queued and while they were running. The RC adopts the tasks and runs them to completion. Upgrading from 4.3.2 to the RC and downgrading back to 4.3.2 also work.
• Make EdgeExecutor respect [core] parallelism #72048: With [core] parallelism = 3, the RC runs at most 3 of 6 Edge tasks at a time on Airflow 3.3.2 and 3.0.6. A slot comes back when a task succeeds, fails and retries, or is marked failed through the REST API.

[shahar1]

Tested edge3 5.0.0rc2 and it works as expected:

• Fix scheduler crash on restart with queued or running Edge tasks #73832: Restarted the scheduler on Airflow 3.3.2 while Edge tasks were queued and while they were running. The RC adopts the tasks and runs them to completion. Upgrading from 4.3.2 to the RC and downgrading back to 4.3.2 also work.
• Make EdgeExecutor respect [core] parallelism #72048: With [core] parallelism = 3, the RC runs at most 3 of 6 Edge tasks at a time on Airflow 3.3.2 and 3.0.6. A slot comes back when a task succeeds, fails and retries, or is marked failed through the REST API.

Thanks!
Please note that Edge provider will be excluded to include #73918, but testing the provider is still welcome.

[wolvery]

Tested PR #69762 in apache-airflow-providers-cncf-kubernetes==10.23.0rc2.

• 32 targeted database-backed tests passed across Airflow 3.3.2 and 3.0.6.
• Live Kubernetes verification using Airflow 3.3.2, PostgreSQL, and two schedulers passed in sequential and async pod-creation modes:
  • No pods were created for workloads with a terminal state, missing/replaced task-instance ID, changed try number, or changed scheduler owner.
  • Fresh workloads created pods and completed successfully in both modes.
  • A normally scheduled Dag completed successfully.

Testing used the published RC, not checkout code. These results verify the fix’s pre-launch stale-workload checks; simultaneous HA races and performance were not tested.

PR #69762 is verified within this scope.

---

Drafted-by: OpenCode (gpt-6.1-sol); reviewed by @wolvery before posting

[anishgirianish]

Tested cncf.kubernetes 10.23.0rc2 and edge3 5.0.0rc2. #73764 and #63491 work as expected. +1 for both.

thanks

[shahar1]

hank you everyone. Providers are released.

I invite everyone to help improve providers for the next release, a list of open issues can be found here.