- Watch requests that are delegated to aggregated API servers no longer reserve concurrency units (seats) in the API Priority and Fairness dispatcher for their entire duration. ([#105511](https://github.com/kubernetes/kubernetes/pull/105511), [@benluddy](https://github.com/benluddy))

*   `tf.config.experimental_list_devices` has been removed, please use
    `tf.config.list_logical_devices`.

## Bug Fixes and Other Changes

*   `tf.data`
    *   Fixes concurrency issue with `tf.data.experimental.parallel_interleave`
        with `sloppy=True`.
    *   Add `tf.data.experimental.dense_to_ragged_batch()`.
    *   Extend `tf.data` parsing ops to support `RaggedTensors`.

    is_body_allowed_for_status_code,
)
from starlette import routing
from starlette._exception_handler import wrap_app_handling_exceptions
from starlette._utils import is_async_callable
from starlette.concurrency import iterate_in_threadpool, run_in_threadpool
from starlette.datastructures import FormData
from starlette.exceptions import HTTPException
from starlette.requests import Request

- Update kube-apiserver's priority & fairness work estimator such that 'max seats' is MIN(0.15 x nominalCL, nominalCL / handSize)
  
  This fixes a bug where clients with requests using hand size x max seats greater than the nominal concurrency limit can starve other requests in the same priority level. ([#118601](https://github.com/kubernetes/kubernetes/pull/118601), [@andrewsykim](https://github.com/andrewsykim)) [SIG API Machinery]

## Dependencies

- Update kube-apiserver's priority & fairness work estimator such that 'max seats' is MIN(0.15 x nominalCL, nominalCL / handSize)
  
  This fixes a bug where clients with requests using hand size x max seats greater than the nominal concurrency limit can starve other requests in the same priority level. ([#118601](https://github.com/kubernetes/kubernetes/pull/118601), [@andrewsykim](https://github.com/andrewsykim)) [SIG API Machinery]

En ajoutant `standard`, Uvicorn va installer et utiliser quelques dépendances supplémentaires recommandées.

Cela inclut `uvloop`, le remplaçant hautes performances de `asyncio`, qui fournit le gros gain de performance en matière de concurrence.

Lorsque vous installez FastAPI avec quelque chose comme `pip install "fastapi[standard]"`, vous obtenez déjà `uvicorn[standard]` aussi.

///

## Exécuter le programme serveur { #run-the-server-program }

* l’en-tête `Authorization`
* l’en-tête transféré

* le système d’injection de dépendances
* la dépendance
* l’élément dépendable
* le dépendant

* lié aux E/S
* lié au processeur
* concurrence
* parallélisme
* multi-traitement

* la variable d’env
* la variable d’environnement
* le `PATH`
* la variable `PATH`

* l’authentification
* le fournisseur d’authentification

### Other (Bug, Cleanup or Flake)

- Azure: fix concurreny issue in lb creation ([#89604](https://github.com/kubernetes/kubernetes/pull/89604), [@aramase](https://github.com/aramase)) [SIG Cloud Provider]

- Azure auth module for kubectl now requests login after refresh token expires. ([#86481](https://github.com/kubernetes/kubernetes/pull/86481), [@tdihp](https://github.com/tdihp)) [SIG API Machinery and Auth]
- Azure: fix concurreny issue in lb creation ([#89604](https://github.com/kubernetes/kubernetes/pull/89604), [@aramase](https://github.com/aramase)) [SIG Cloud Provider]