Status of testing Providers that were prepared on May 19, 2026

[jscheffl]

Body

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 airbyte: 5.5.0rc1

• Implement execution_timeout semantics for AirbyteTriggerSyncOperator in deferrable mode (#64051): @SameerMesiah97
  Linked issues:
  • Linked Issue #61472: @SameerMesiah97

Provider alibaba: 3.3.8rc1

• Fix Alibaba OSS provider: configurable endpoint and task handler log … (#66479): @hasangzc

Provider amazon: 9.29.0rc1

• Add OpenLineage support to AthenaSQLHook (#66844): @rahul-madaan
• Add S3TablesDeleteTableBucketPolicyOperator (#66957): @john-jac
• Add OpenLineage parent and transport info injection to EmrContainerOperator (#66816): @rahul-madaan
  Linked issues:
  • Linked Issue #64513: @rahul-madaan
  • Linked Issue #64807: @rahul-madaan
• Add MwaaServerlessDeleteWorkflowOperator (#66891): @kars0508
• feat: add callback support to aws batch executor (#62984): @dondaum
  Linked issues:
  • Linked Issue #62887: @ferruzzi
• Add MwaaServerlessUpdateWorkflowOperator (#66833): @kars0508
• Add S3TablesPutTableBucketPolicyOperator (#66720): @john-jac
• Add S3VectorsQueryVectorsOperator (#66724): @john-jac
• Adds async support to SageMakerNotebookJobTrigger (#65571): @Cloud-Architect-Emma
• Add S3VectorsPutVectorsOperator (#66545): @john-jac
• Add S3TablesRenameTableOperator (#66544): @john-jac
• Add MwaaServerlessStopWorkflowRunOperator (#66548): @john-jac
• Add BedrockUpdateGuardrailOperator (#66547): @john-jac
• Add GlueCatalogCreatePartitionOperator (#66546): @john-jac
• Add MwaaServerlessWorkflowRunSensor (#66382): @john-jac
• Add BedrockCreateEvaluationJobOperator (#66722): @john-jac
• Fix CreateWorkflowOperator if_exists=skip ARN construction (#66974): @kars0508
• Fix EMR container job not cancelled on deferral timeout (#64770): @aurangzaib048
  Linked issues:
  • Linked Issue #51883: @dominikhei
  • Linked Issue #60517: @arorasachin9
• Add Bundle documentation for Amazon S3 and Google GCS providers (#66993): @shashbha14
  Linked issues:
  • Linked Issue #66986: @eladkal

Provider apache.kafka: 1.14.0rc1

• Fix KafkaError.name() called without parentheses in create_topic (#65734): @david-parkk
• Fix AwaitMessageTrigger missing _task_instance attribute (#66302): @jason810496
  Linked issues:
  • Linked Issue #64833: @jason810496

Provider apache.livy: 4.5.6rc1

• Include endpoint_prefix in LivyTrigger serialization (#66966): @Nishieee
  Linked issues:
  • Linked Issue #66961: @shahar1

Provider apache.spark: 6.0.2rc1

• Fix SparkPipelinesHook for Spark Connect connections (#66498): @lisancao

Provider celery: 3.20.0rc1

• feat: add callback support to aws batch executor (#62984): @dondaum
  Linked issues:
  • Linked Issue #62887: @ferruzzi

Provider cncf.kubernetes: 10.17.1rc1

• Fix deferrable KPO trigger_reentry crash when pod is GC'd before re-entry (#66716): @jekeanyanwu
  Linked issues:
  • Linked Issue #56976: @m1racoli
  • Linked Issue #66715: @jekeanyanwu
  • Linked Issue #39296: @antoniocorralsierra
• fix(k8s-executor): add latency + status metrics around pod API calls (#66806): @1fanwang
• Re-defer KubernetesPodOperator when pod is still running after trigger error (#66705): @wolfdn
• Fix KubernetesPodTrigger pod terminal state handling (#66650): @anmolxlight
• Fix race condition in AsyncKubernetesHook corrupting global kubernetes_asyncio config (#65566): @ZhaoMJ
• fix/issue-66592: Respecting unwrap_single for non-deferrable execution (#66596): @jroachgolf84
  Linked issues:
  • Linked Issue #66592: @farrukh-t
• Fix kubernetes cleanup-pods ignoring --verbose (#65955): @henry3260
• Preserve event resource version after empty sync polls (#66471): @hkc-8010
  Linked issues:
  • Linked Issue #66470: @hkc-8010
• Fix SparkKubernetesOperator with deferrable=True falling through to KubernetesPodOperator.execute() creating a spurious pod (#66448): @Pranaykarvi
  Linked issues:
  • Linked Issue #66437: @nikhilkarve
• KubernetesExecutor: scope periodic completed-pod adoption to dead schedulers (#66400): @potiuk
  Linked issues:
  • Linked Issue #66396: @potiuk
  • Linked Issue #61839: @atrbgithub

Provider common.ai: 0.3.0rc1

• Add LLMRetryPolicy to common-ai provider (#65451): @kaxil
  Linked issues:
  • Linked Issue #65474: @kaxil
• fix: update dependencies to fix dependabot alarms in providers.common.ai (#66628): @jscheffl

Provider common.compat: 1.15.0rc1

• Add BundleVersion dataclass and version_data persistence to DagVersion (#66491): @o-nikolas

Provider common.sql: 2.0.0rc1

• feat: Add standardized SQL check representation for listeners (#66849): @kacpermuda

Provider databricks: 7.15.0rc1

• Databricks: support user-assigned managed identity for Azure VM auth (#66072): @sydongda
  Linked issues:
  • Linked Issue #65588: @sydongda
• fix issue: #64828, repair missing job parameters if exists (#67055): @sydongda
  Linked issues:
  • Linked Issue #64828: @roydobbe
• Preserve Databricks deferrable trigger caller across triggerer restarts (#66965): @Nishieee
  Linked issues:
  • Linked Issue #66961: @shahar1
• Fix DatabricksWorkflowTaskGroup leaking TaskGroupContext on internal exception (#66582): @moomindani
  Linked issues:
  • Linked Issue #55891: @ayudovin
  • Linked Issue #42164: @fedemgp

Provider dbt.cloud: 4.9.0rc1

• Add DbtCloudListJobRunsOperator to DBT Cloud Provider (#66150): @SameerMesiah97
• Refactor deferrable execution in DbtCloudRunJobOperator (#66449): @SameerMesiah97
  Linked issues:
  • Linked Issue #64051: @SameerMesiah97

Provider docker: 4.5.6rc1

• Fix DockerOperator init crash on dict mount entries (#66553): @bdsoha
  Linked issues:
  • Linked Issue #52451: @bdsoha

Provider edge3: 3.7.0rc1

• Allow using fresh interpreter besides fork() in Edge Worker (#65943): @diogosilva30
  Linked issues:
  • Linked Issue #65942: @diogosilva30
• Support BaseExecutor.run_workload() for Airflow 3.3 in Edge Worker (#65847): @jscheffl
• Bugfix/make edge resilient against log errors (#66995): @jscheffl
• Fix provider DB upgrades with existing tables (#66883): @anmolxlight
  Linked issues:
  • Linked Issue #66674: @anmolxlight
• Refresh uv.lock and Edge OpenAPI spec after #67093 env upgrade (#67188): @vatsrahul1001
  Linked issues:
  • Linked Issue #67174: @jason810496
  • Linked Issue #67180: @zachliu
  • Linked Issue #66574: @potiuk
  • Linked Issue #67184: @Lee-W
  • Linked Issue #67093: @github-actions[bot]
  • Linked Issue #67130: @bugraoz93

Provider elasticsearch: 6.5.4rc1

• ElasticsearchRemoteLogIO: Harden _parse_raw_log to handle malformed and non-JSON log lines (#66383): @SameerMesiah97
• fix(elasticsearch): pin compatible-with at the transport layer to keep ES 8 servers working (#66065): @petercheon
  Linked issues:
  • Linked Issue #66064: @petercheon
  • Linked Issue #64070: @Subham-KRLX
• ElasticsearchSQLCursor: Implement fetchmany support (#66658): @SameerMesiah97

Provider fab: 3.6.4rc1

• Fix provider DB upgrades with existing tables (#66883): @anmolxlight
  Linked issues:
  • Linked Issue #66674: @anmolxlight
• Fix FAB role deletion foreign key constraint violation (#66686): @kimjune01
  Linked issues:
  • Linked Issue #65375: @TKul6
• Fix session cookie serialization when using securecookie backend (#66565): @smartlin5228
• fix: the ldap authentication handler in the flask-ap... in override.py (#66417): @orbisai0security
• Bump flask-appbuilder to 5.2.1 and mirror new auth event hooks (#66841): @potiuk
  Linked issues:
  • Linked Issue #2440: @abhijeetdhumal
  • Linked Issue #66840: @potiuk
  • Linked Issue #2450: @mdziemianko
  • Linked Issue #2451: @lvicentesanchez
• Pin pyjwt>=2.11.0 in FAB provider so flask_jwt_extended imports cleanly (#66840): @potiuk
  Linked issues:
  • Linked Issue #61278: @amoghrajesh

Provider git: 0.4.0rc1

• Add Git Sparse Checkout to Git Dag Bundle (#67047): @jscheffl
• Add BundleVersion dataclass and version_data persistence to DagVersion (#66491): @o-nikolas

Provider google: 22.0.0rc1

• Remove MLEngine from google provider (#66526): @Crowiant
• Add host_key_policy option to ComputeEngineSSHHook (#66746): @potiuk
• Add BigQueryStreamingBufferEmptySensor for DML on streaming tables (#66652): @radhwene
  Linked issues:
  • Linked Issue #59408: @radhwene
  • Linked Issue #61148: @radhwene
• Delete unused job_id_path XCom push from BigQueryInsertJobOperator (#66760): @VladaZakharova
• Migrate BigQueryInsertJobTrigger to use on_kill() for user-initiated kills (#66704): @kir-rin
  Linked issues:
  • Linked Issue #65733: @amoghrajesh
  • Linked Issue #65590: @amoghrajesh
• Add Cloud SQL Auth Proxy IAM authentication (#66510): @nailo2c
  Linked issues:
  • Linked Issue #35347: @jitendra390
• Add aliases for rebranded Google services (#66344): @VladaZakharova
• Add BigQuery routine operators and existence sensor (#65499): @alamashir
  Linked issues:
  • Linked Issue #65467: @alamashir
  • Linked Issue #47579: @luke-hoffman1
• Skip S3 folder-marker keys in S3ToGCSOperator (#65724): @yuseok89
• Support all bq load job and ext table config options in GCSToBigQueryOperator (#64505): @mlauter
• Preserve BigQueryIntervalCheckTrigger params after triggerer restart (#67053): @Vamsi-klu
  Linked issues:
  • Linked Issue #66961: @shahar1
• Fix CloudRunExecuteJobOperator deferrable mode silently passing on cancel (#67050): @Vamsi-klu
  Linked issues:
  • Linked Issue #57791: @MarcusCramer91
• Render Jinja templates in CloudBatchSubmitJobOperator job field (#67021): @nailo2c
  Linked issues:
  • Linked Issue #37217: @ya7010
• Adapt GCP CloudSQL trigger to run in private cloud (#66917): @olegkachur-e
• Serialize poll_interval and impersonation_chain on DataFusionStartPipelineTrigger (#66968): @Nishieee
  Linked issues:
  • Linked Issue #66961: @shahar1
• Fix BigQueryStreamingBufferEmptySensor failing against real BigQuery (#66962): @shahar1
  Linked issues:
  • Linked Issue #66963: @shahar1
  • Linked Issue #66652: @radhwene
• Fix CloudSecretManagerBackend regression with explicit project_id (#61654): @dv-gorasiya
  Linked issues:
  • Linked Issue #61217: @kiranpreet-kaur17
• Fix broken GKEStartPodOperator extra link (#66657): @Vamsi-klu
  Linked issues:
  • Linked Issue #46658: @mdering
• Adjust GCP BigQuery triggers for the private cloud (#66404): @olegkachur-e
• Add Bundle documentation for Amazon S3 and Google GCS providers (#66993): @shashbha14
  Linked issues:
  • Linked Issue #66986: @eladkal
• Document Dataproc limitations in Sovereign Cloud from Google (#66706): @VladaZakharova

Provider imap: 3.11.3rc1

• Neutralize path separator in IMAP (#66951): @jscheffl
• ImapHook: Handle mixed plaintext and RFC 2047 encoded attachment filenames (#66672): @SameerMesiah97
  Linked issues:
  • Linked Issue #66578: @mgustimz
  • Linked Issue #65875: @MdSadiqMd
• fix RFC 2047 filename decoding (#65875): @MdSadiqMd
  Linked issues:
  • Linked Issue #65871: @vaefremov95

Provider influxdb: 2.11.0rc1

• Add support for influx3 in influxdb provider (#58929): @arpitrathore
  Linked issues:
  • Linked Issue #58610: @arpitrathore

Provider keycloak: 0.7.2rc1

• Fix Keycloak default role policy wiring (#67031): @anmolxlight
  Linked issues:
  • Linked Issue #67027: @mogilevskiikonstantin-commits
• Update Keycloak documentation to mention [api] base_url (#66023): @vincbeck
  Linked issues:
  • Linked Issue #65952: @n-badtke-cg

Provider microsoft.azure: 13.3.0rc1

• Add deferrable mode to AzureContainerInstancesOperator (#62772): @cruseakshay
  Linked issues:
  • Linked Issue #62433: @tsanton
• Fix Azure Batch provider import error by capping azure-batch<15 (#66452): @wjddn279
• Fix error messages in PythonVirtualenvOperator when Azure Key Vault secret backend is configured (#67157): @wolfdn

Provider openlineage: 2.17.0rc1

• feat: Add standardized SQL check representation for listeners (#66849): @kacpermuda
• cache connections in OpenLineage SQL hook lineage (#64843): @mobuchowski

Provider pinecone: 2.4.5rc1

• Stop passing project_id to the Pinecone client constructor (#66601): @potiuk

Provider sftp: 5.8.0rc1

• SFTP: add async retrieve_file / store_file / mkdir to SFTPHookAsync and introduce SFTPClientPool (#64465): @dabla

Provider smtp: 3.0.1rc1

• Fix smtp connection test for noop tuple response (#66406): @henry3260
  Linked issues:
  • Linked Issue #66391: @omacchioni
• fix stmp email callback to use email_conn_id (#65072): @wjddn279

Provider snowflake: 6.13.0rc1

• Added statement timeout support in SnowflakeSqlApiHook & Operator (#63575): @justinpakzad
  Linked issues:
  • Linked Issue #60894: @HyunWooZZ
• Replace AirflowExceptions with native exceptions in Snowflake provider (#66279): @justinpakzad
• Add PAT connection to snowflake in SnowflakeHook docs (#66953): @romsharon98
• Add Snowflake data quality operator examples to provider docs and system tests (#65623): @PrithviBadiga
  Linked issues:
  • Linked Issue #65098: @nagabharathnr

Provider standard: 1.13.1rc1

• fix: Handle None Task ID state when matching states in ExternalTaskSensor (#66881): @jroachgolf84
  Linked issues:
  • Linked Issue #66877: @nathadfield
• Fix error messages in PythonVirtualenvOperator when Azure Key Vault secret backend is configured (#67157): @wolfdn

Provider tableau: 5.4.1rc1

• Fix TableauOperator tasks.run AttributeError (#66682): @Vamsi-klu
  Linked issues:
  • Linked Issue #42248: @farrukh-t

Provider weaviate: 3.3.4rc1

• [main] Upgrade important CI environment (#66700): @github-actions[bot]

Provider zendesk: 4.12.0rc1

• Support API authentication for Zendesk (#64591): @Subham-KRLX
  Linked issues:
  • Linked Issue #64589: @fjmacagno

All users involved in the PRs:
@hasangzc @anmolxlight @PrithviBadiga @dabla @john-jac @jscheffl @moomindani @smartlin5228 @vatsrahul1001 @henry3260 @diogosilva30 @Pranaykarvi @Nishieee @kaxil @sydongda @jason810496 @kacpermuda @bdsoha @cruseakshay @shahar1 @mobuchowski @justinpakzad @kars0508 @wolfdn @aurangzaib048 @1fanwang @Vamsi-klu @Subham-KRLX @hkc-8010 @petercheon @kimjune01 @orbisai0security @vincbeck @nailo2c @dv-gorasiya @SameerMesiah97 @romsharon98 @MdSadiqMd @olegkachur-e @radhwe

Committer

• I acknowledge that I am a maintainer/committer of the Apache Airflow project.

[petercheon]

Verified #66065 — fix(elasticsearch): pin compatible-with at the transport layer to keep ES 8 servers working in apache-airflow-providers-elasticsearch==6.5.4rc1. ✅

Environment

• Airflow 3.1.3 (clean venv, Python 3.11)
• apache-airflow-providers-elasticsearch==6.5.4rc1 from PyPI
• elasticsearch-py==9.4.0 (the default >=9 resolution that triggers the regression)
• Server: live Elasticsearch 8.17.3 cluster

Steps

1. Reproduced the regression first (no fix applied): a raw Elasticsearch(...) client against the ES 8 server fails with the documented error —

   BadRequestError(400, 'media_type_header_exception',
     Accept version must be either version 8 or 7, but found 9.
     Accept=application/vnd.elasticsearch+json; compatible-with=9)
   

2. Applied the fix by setting AIRFLOW__ELASTICSEARCH__ES_COMPAT_WITH=8 and calling airflow.providers.elasticsearch._compat.apply_compat_with(client). The following all succeed against the same ES 8 server:

   • client.info() → version.number = 8.17.3
   • client.cluster.health() → status=green
   • client.search(index="missing") → clean 404 (no media-type error)
   • client.bulk(operations=...) → errors=False, both docs indexed, count=2

3. Wire-level confirmation — swapped elastic_transport.Transport.perform_request for a spy and captured the headers actually sent:

   info()  Accept       : application/vnd.elasticsearch+json;    compatible-with=8
   bulk()  Accept       : application/vnd.elasticsearch+json;    compatible-with=8
   bulk()  Content-Type : application/vnd.elasticsearch+x-ndjson; compatible-with=8
   

   The +x-ndjson Content-Type rewrite (the originally-broken bulk path) is correctly applied.

4. End-to-end through the hook: ElasticsearchPythonHook returns a client whose transport is wrapped when es_compat_with is set, and unwrapped when it isn't — same hook.search(...) call succeeds with the option set and fails with the 400 without it.

Looks good to ship from my side. 🎉

[Subham-KRLX]

Verified both Zendesk 4.12.0rc1 and Elasticsearch 6.5.4rc1. I ran the Zendesk unit tests covering the new API and OAuth token authentication paths and the UI connection behavior and they passed successfully. I also ran the Elasticsearch unit tests covering log ingestion index patterns and client compatibility and they all passed. Both look good to ship.

[Nishieee]

Hey!! I've verified the RC builds for my provider changes (installed from PyPI, tested via .venv imports from site-packages):

• apache-airflow-providers-apache-livy==4.5.6rc1 (#66966): LivyTrigger preserves endpoint_prefix through serialize() and restore.
• apache-airflow-providers-databricks==7.15.0rc1 (#66965): caller round-trips on DatabricksExecutionTrigger and DatabricksSQLStatementExecutionTrigger.
• apache-airflow-providers-google==22.0.0rc1 (#66968): poll_interval and impersonation_chain round-trip on DataFusionStartPipelineTrigger.

I don't have Livy, Databricks, or Data Fusion staging set up right now, so I couldn't run a full deferrable + triggerer-restart smoke test but the serialization checks against the RC wheels all passed on my side. If deeper integration testing would help, I'm glad to stand up environments and run those. just let me know.

Thanks!

[david-parkk]

Verified apache-airflow-providers-apache-kafka==1.14.0rc1 (#65734)

Environment

• Airflow 3.2.1, Python 3.10, PostgreSQL
• apache-airflow-providers-apache-kafka==1.14.0rc1 from PyPI

Steps

• Installed the RC in Breeze and ran kafka_exception_inspect DAG against a live Kafka broker — specifically to verify the KafkaError.name() fix I contributed (Fix KafkaError.name() called without parentheses in create_topic #65734).
• Also run the unit tests (pytest providers/apache/kafka/tests/unit/apache/kafka/hooks/test_client.py -xvs)

thank you 😄

[dondaum]

#62984 looks good.
I have downloaded both RC versions for the amazon and celery providers. I can confirm that all relevant code changes are present.

[Cloud-Architect-Emma]

Verified apache-airflow-providers-amazon==9.29.0rc1 (#65571)

I verified that the SageMakerNotebookJobTrigger async implementation
is present and correct in the RC wheel:

• SageMakerNotebookJobTrigger.run() is now a proper async generator
  using asyncio.to_thread for CI-safe async execution
• serialize() correctly round-trips all four fields: execution_id,
  execution_name, waiter_delay, waiter_max_attempts
• Terminal success states (COMPLETED, SUCCEEDED) and failure states
  (FAILED, ERROR, CANCELLED, STOPPED) are handled correctly
• Timeout behavior yields a TriggerEvent with status: "failed" and
  error: "Execution timed out"

The implementation and all unit tests are present in the RC package.

[moomindani]

Tested apache-airflow-providers-databricks==7.15.0rc1 for PR #66582 (DatabricksWorkflowTaskGroup TaskGroupContext leak fix, closing #42164) — verification passes.

Installed the RC wheel into a fresh apache-airflow==3.1.3 venv and ran the original reproducer from #42164: a DatabricksWorkflowTaskGroup with an EmptyOperator inside (which lacks _convert_to_databricks_workflow_task and therefore makes the workflow group's __exit__ raise mid-way). With the RC installed:

• the expected AirflowException: Task wf.not_convertible does not support conversion to databricks workflow task. is raised,
• TaskGroupContext._context is empty afterwards (no leak), and
• a subsequent DatabricksWorkflowTaskGroup builds cleanly without the Cannot mix TaskGroups from different DAGs error that was the user-visible symptom on unfixed code.

Marking my line in the Databricks section.

[kacpermuda]

@jscheffl Not that I have anything against, but why are we doing a major release of common.sql? I see only my feature PR since last release, and it's definitely not a breaking change.

[kacpermuda]

Both my PRs are ok, tested my changes

[sydongda]

Tested RC: apache-airflow-providers-databricks==7.15.0rc1
Related PRs: #66072, #67055

Test steps

1. Upgraded provider to 7.15.0rc1
2. Created a Databricks connection using user managed identity auth
3. Created a DAG using the job_parameters param
4. First run – with job_parameters
   • Checkpoint 1: connection to Databricks → ✅ pass
   • Checkpoint 2: repair run with job_parameters → ✅ pass
5. Second run – without job_parameters
   • Checkpoint: repair run without job_parameters → ✅ pass

Conclusion

✅ Both PRs (#66072, #67055) are addressed.
✅ No regressions observed for the tested scenarios.

[VladaZakharova]

Google provider passed the tests
+1 for Google changes

[zachliu]

#67180 passed the tests 🚀

[jscheffl]

@jscheffl Not that I have anything against, but why are we doing a major release of common.sql? I see only my feature PR since last release, and it's definitely not a breaking change.

Oh, I needed a moment to triage backwards "how" this came. After a good sleep and a hard working day my memory is a bit dissolved from what I did yesterday but mechanics is as you nee in the published release notes of RC in https://airflow.staged.apache.org/docs/apache-airflow-providers-common-sql/2.0.0/changelog.html that there is a big fat warning banner about breaking change. This was in front of the changelog which by release manager mechanic the signal that a breaking change was added and this caused the major version bump.

But as you hinted... looking at the history from past providers release from 2026-05-05 in https://github.com/apache/airflow/pull/66424/changes#diff-67507f2803366fd3ff09a384dcb696bfd4a20ab1e75bb3e15f289c63e37b18edR28 I need to acknowledge that re-reading the old state that by accident the braking change was actually missed 2 weeks ago in last providers release 1.36.0 - actually this would have needed to be 2.0.0.

Due to the changelog and version mis-match... I assume it is fair to retro-major-bump the provider so that everybody in <= 1.36 sees there is a breaking change with warning. Would have been better last release cycle and this today we would have 2.1.0.

Hope it is okay to keep 2.0.0 :-D

(FYI @vincbeck)