…t_authorized_variables` in Fab auth manager (#55682)

closes: #53817

…ves extra content, optimizes description (#55564)

* feat(i18n): Add Chinese (zh-CN) internationalization support

This commit adds Simplified Chinese (zh-CN) internationalization support to Airflow UI.

The changes include:
- Added complete Simplified Chinese translation files
- Updated i18n configuration to include zh-CN locale
- Updated translation completeness check script
- Updated CODEOWNERS for zh-CN translations
- Optimized translations for better user experience
- Unified terminology for consistency (e.g. backfill, login)
- Updated DAG/DAGs capitalization according to #55099

Co-authored-by: @gyli

* Apply suggestions from code review

* fixup! Apply suggestions from code review

---------

Co-authored-by: Kaxil Naik <kaxilnaik@gmail.com>

There are various processes in Airflow that are not yet team-ified and
that rely on executors (serving logs, callback handling) so for now,
enforce that at least one global executor is present to be used for such
things. This can be revisited and relaxed at a future time.

This updates screenshots in core concepts – dags because of UI refresh

Refs: #55521

)

Resolves an issue where DAGs using non default value like `weight_rule='absolute'` would fail during
deserialization with `'Unknown priority strategy'` error. The problem occurred
when the scheduler attempted to deserialize tasks that had already been
converted to `PriorityWeightStrategy` instances.

The `validate_and_load_priority_weight_strategy` function now properly handles
cases where a `PriorityWeightStrategy` instance is passed, returning it directly
instead of attempting to re-validate it.

This fix ensures that all weight rule options (`downstream`, `upstream`, `absolute`)
work correctly in the complete DAG processing pipeline including serialization,
storage, and deserialization.

To Reproduce the bug, run the following dag or just run the added unit test without code changes:

```py
from airflow import DAG
from airflow.sdk import task
from airflow.task.weight_rule import WeightRule

with DAG("xcom_backend_dag"):

    @task()
    def xcom_producer():
        return "my_value"

    @task(weight_rule=WeightRule.ABSOLUTE)
    def xcom_consumer(value):
        assert value == "my_value"

    xcom_consumer(xcom_producer())

```

which will then fail, when you try to trigger the dag from UI.

Fix DAG deserialization failure with weight_rule='absolute'

Resolves an issue where DAGs using weight_rule='absolute' would fail during
deserialization with 'Unknown priority strategy' error. The problem occurred
when the SerializedBaseOperator.weight_rule property was re-validating
already-instantiated PriorityWeightStrategy instances.

During deserialization, decode_priority_weight_strategy() creates strategy
instances and stores them in _weight_rule, but the weight_rule property
was calling validate_and_load_priority_weight_strategy() which expected
a string or class, not an instance.

Fixed by checking if _weight_rule is already a PriorityWeightStrategy
instance in the property getter and returning it directly. User input
still goes through proper validation.

This targeted fix resolves the specific deserialization issue while
maintaining existing validation behavior for new operator creation.

Everything between the 2 mentions of the word "Notifier" in the
"Templating and Context" section was being rendered as inline code
because I made the first one plural (with an "s" right after the double
backticks).

I've fixed this by making Notifier render as regular text instead. This
also makes it more consistent with the rest of the doc page.

…52839)

* Fix: Use get instead of hasattr for task_result in BulkStateFetcher

* add type annotation for task_results_by_task_id

* add type annotation for params in methods of BulkStateFetcher

* retain type annotation only in param level

* add mock value for sync_parallelism in test

* update airflow_version in asset_watcher migration

* handle version compat

* feat(ui): add DAG run state filter to grid view options

  Add filtering by DAG run state (queued, running, success, failed) to the grid view options panel, following the same pattern as the
  existing DagRuns.tsx implementation.

* Add testsgit add .

* Gantt chart filtering issue by applying the same pattern as PR #55845

The issue was that the Gantt chart wasn't using the same filters as the Grid view. When you set a DAG run state filter in the Options
  menu, it would filter the Grid view but not the Gantt chart. Now both views will be synchronized with the same filtering criteria.

* build(open-api): regnerate open api spec

* Fix static checks in Edge

* Fix static checks in Edge v2

---------

Co-authored-by: Wei Lee <weilee.rx@gmail.com>

…ent (#55872)

…and `filter_authorized_variables` in AWS auth manager (#55687)

This commit adds more tests to cover various aspects of DB manager
upgrade and downgrades.

* refmt asset name with fixedWidth hidden & ellipsis

* Update airflow-core/src/airflow/ui/src/pages/DagsList/AssetSchedule.tsx

Co-authored-by: Guan Ming(Wesley) Chiu <105915352+guan404ming@users.noreply.github.com>

* lintg

---------

Co-authored-by: Guan Ming(Wesley) Chiu <105915352+guan404ming@users.noreply.github.com>

This page was difficult to find - even for existing users who are familiar with the layout.

#55878)

Most of them don't need it on either Airflow 2 or 3 - the only exception to
that is `airflow celery worker` which probably does still need it for Airflow
2, as tasks there directly access the DB.

If bad configuration was provided (user tried to configure an executor
for a team that does not exist) then throw an exception and do not allow
execution loading to complete (and Airflow Scheduler to start).

* Add executor, hostname, and queue columns to TaskInstances page

  - Add executor column to display task instance executor information
  - Add hostname column to show where task instance ran
  - Add queue column to display task instance queue assignment

* Disable new columns by default

* Add version_compat for HTTP provider

* Create HTTPEventTrigger class with constructor

* First draft of HttpEventTrigger

* Remove trailing whitespace

* Pass response_check path instead of callable for serialization

* Create unit tests for HttpEventTrigger

* First draft of HttpEventTrigger docs

* Add version_compat for HTTP provider

* Create HTTPEventTrigger class with constructor

* Change log formatting

* Add parameters to docs

* Ruff formatting

* Ruff formatting

* Add check to ensure that response_check is async

* Correct failing tests

* Change imports for Airflow 2 compatibility tests

* Add usage example in docs

* Add usage example to docs

* Add comment about token to docs

* Modify parameters desc

* Make response_check_path mandatory

* Add comments to docs

* Change notes order

* Make _import_from_response_check_path async

* Modify imports

* Correction in docs

* Remove annotations import

* Fix log text selection contrast in light mode

  Changed selection background from brand.subtle to gray.emphasized
  to improve readability when highlighting text in the log viewer.
  This aligns with the DAG code viewer styling and provides better
  contrast across both light and dark themes.

* Brent's suggestions

…om failing and modify example_cloud_logging_sink (#55939)

Co-authored-by: Anton Nitochkin <nitochkin@google.com>

…t support (#56240)

* EdgeExecutor was not implementing the revoke_task() method, causing it to
  raise NotImplementedError when the scheduler attempted to handle tasks stuck
  in queued state. This meant the task_queued_timeout feature was completely
  bypassed for EdgeExecutor.

  This commit implements revoke_task() to:
  - Remove tasks from executor's internal state (running, queued_tasks, last_reported_state)
  - Delete corresponding EdgeJobModel records to prevent edge workers from picking them up
  - Enable proper task queued timeout handling by the scheduler

  Added tests to verify revoke_task correctly removes tasks from both executor
  state and database, and handles non-existent tasks gracefully.

* Update providers/edge3/src/airflow/providers/edge3/executors/edge_executor.py

Co-authored-by: Jens Scheffler <95105677+jscheffl@users.noreply.github.com>

---------

Co-authored-by: Jens Scheffler <95105677+jscheffl@users.noreply.github.com>

* Add async support for Amazon SQS Notifier

* fix bad test

MESSAGE_ID is not the same is MESSAGE_BODY, that assertion was failing as it should have.  Removed the bad assertion

* Prepare fab and amazon providers to release (September 2025)

* UI: Add Expand/Collapse all to XComs page

* Restored few changes

* Fixing lint issue

* Made suggested Changes

…56266)

These imports ultimately aren't cheap - it results in ~80mb of memory
use, so lets skip it unless the plugin is enabled.

…of wait_for_completion (#56158)

* refactor: deprecate wait_policy in EmrCreateJobFlowOperator in favor of wait_for_completion

* added unit test for refactor

* resolved copilot comments

* resolved copilot comments

* fixed failing test

* fixed: refactor of wait_policy

* ensured backward compatibility

* removed "self.wait_policy = wait_policy"

Ensure removed/historical tasks from selected runs are visible in
Grid even if they no longer exist in the current DAG version.

We now:
- Include synthetic leaf nodes for task_ids present in TIs but
missing from the serialized DAG in both grid/structure and grid/ti_summaries.
- Aggregate TI states for these synthetic nodes

Add tests covering structure and TI summaries for removed tasks.

…#56296)

This reverts commit 3150430.

Co-authored-by: Brent Bovenzi <brent@astronomer.io>

As the Edeg API server is currently implemented we need to "pretend" to be  a
specific Airflow and Edge provider version. These default to the currently
released versions, but can be changed via env vars to work elsewhere.

This works enough to run tasks, but there might need to be some changes to the
Edge API to support non-python clients (for example, working out the
versioning strategy to make it long-term supportible and not need the Airflow
and Edge Provider version to match 100%).

A chunk of the changes here are to make the config and global setup "more well
structured" -- so that they are suitable to be easily called from multiple
workers (Celery and Go, useful if we don't end up keeping the Celery worker)

* modify test_ftp

Signed-off-by: Xch1 <qchwan@gmail.com>

* modify test_sql

Signed-off-by: Xch1 <qchwan@gmail.com>

* modify test_exasol

Signed-off-by: Xch1 <qchwan@gmail.com>

* modify test_es_task_handler

Signed-off-by: Xch1 <qchwan@gmail.com>

* modify test_edge_executor

Signed-off-by: Xch1 <qchwan@gmail.com>

* fix test_edge_executor match

Signed-off-by: Xch1 <qchwan@gmail.com>

* fix test_edge_executor match

Signed-off-by: Xch1 <qchwan@gmail.com>

* fix test_edge_executor match

Signed-off-by: Xch1 <qchwan@gmail.com>

---------

Signed-off-by: Xch1 <qchwan@gmail.com>

)

EKS test sometimes fails with exceptions like:
`botocore.errorfactory.ResourceInUseException: An error occurred (ResourceInUseException) when calling the DeleteFargateProfile operation: Cannot Delete Fargate Profile [profile name] because cluster [cluster name] currently has an update in progress`

This is (a potentially temporary) fix for that error.

* Use TI duration from db instead of UI calculated

* Fix barchart heights

* Avoid using rem for icons for safari compatibility

* Fix poorly sized custom icon