and it could end up Pods rejected by Extenders stuck in unschedulable pod pool in 5min in the worst-case scenario.
  Now, the scheduling queue notices extenders' failures and requeue Pods rejected by Extenders appropriately. ([#122022](https://github.com/kubernetes/kubernetes/pull/122022), [@sanposhiho](https://github.com/sanposhiho)) [SIG Scheduling]

### Other (Cleanup or Flake)

    * Update the metadata agent's liveness probe to a new /healthz handler.
    * Logging Agent Improvements
    * Bump logging agent version to 0.2-1.5.33-1-k8s-1.
    * Appropriately set log severity for k8s_container.
    * Fix detect exceptions plugin to analyze message field instead of log field.
    * Fix detect exceptions plugin to analyze streams based on local resource id.

* Scheduler metrics are now marked as with the ALPHA stability level. ([#81576](https://github.com/kubernetes/kubernetes/pull/81576), [@logicalhan](https://github.com/logicalhan))
* cache-control headers are now set appropriately.  Only openapi is cacheable if etags match. ([#81946](https://github.com/kubernetes/kubernetes/pull/81946), [@deads2k](https://github.com/deads2k))

        [guide](https://www.tensorflow.org/guide/concrete_function) for more
        details on `concrete_ function`.
    *   Update `tf.function`'s `experimental_relax_shapes` to handle composite
        tensors appropriately.
    *   Optimize `tf.function` invocation, by removing redundant list converter.
    *   `tf.function` will retrace when called with a different variable instead
        of simply using the `dtype` & `shape`.