- Allowed scheduling framework plugins that implement `io.Closer` to be gracefully closed.
   ([#122498](https://github.com/kubernetes/kubernetes/pull/122498), [@Gekko0114](https://github.com/Gekko0114))
- Bumped cAdvisor to `v0.49.0`.
   ([#123599](https://github.com/kubernetes/kubernetes/pull/123599), [@bobbypage](https://github.com/bobbypage))

## Changelog since v1.4.0-beta.3

### Other notable changes

* Bumped Heapster to v1.2.0. ([#32649](https://github.com/kubernetes/kubernetes/pull/32649), [@piosz](https://github.com/piosz))
    * More details about the release https://github.com/kubernetes/heapster/releases/tag/v1.2.0

* Docker digest validation is too strict ([#32627](https://github.com/kubernetes/kubernetes/pull/32627), [@smarterclayton](https://github.com/smarterclayton))
* Bumped Heapster to v1.2.0. ([#32649](https://github.com/kubernetes/kubernetes/pull/32649), [@piosz](https://github.com/piosz))
    * More details about the release https://github.com/kubernetes/heapster/releases/tag/v1.2.0

      return false;
    }

    // This implementation is patterned after ConcurrentHashMap, but without the locking. The only
    // way for it to return a false negative would be for the target value to jump around in the map
    // such that none of the subsequent iterations observed it, despite the fact that at every point
    // in time it was present somewhere int the map. This becomes increasingly unlikely as

* Fluentd/Elastisearch add-on: correctly parse and index kubernetes labels ([#36857](https://github.com/kubernetes/kubernetes/pull/36857), [@Shrugs](https://github.com/Shrugs))

### Heapster
* Bumped Heapster to v1.3.0. ([#43298](https://github.com/kubernetes/kubernetes/pull/43298), [@piosz](https://github.com/piosz))
  * More details about the release https://github.com/kubernetes/heapster/releases/tag/v1.3.0

### Registry

- Bumped addon manager to use `kubectl` version `v1.32.2`. ([#130548](https://github.com/kubernetes/kubernetes/pull/130548), [@Jefftree](https://github.com/Jefftree)) [SIG Cloud Provider, Scalability and Testing]

* Ensure euqiv hash calculation is per schedule ([#59245](https://github.com/kubernetes/kubernetes/pull/59245), [@resouer](https://github.com/resouer))

* Upped the timeout for apiserver communication in the juju kubernetes-worker charm. ([#59219](https://github.com/kubernetes/kubernetes/pull/59219), [@hyperbolic2346](https://github.com/hyperbolic2346))

      return false;
    }

    // This implementation is patterned after ConcurrentHashMap, but without the locking. The only
    // way for it to return a false negative would be for the target value to jump around in the map
    // such that none of the subsequent iterations observed it, despite the fact that at every point
    // in time it was present somewhere int the map. This becomes increasingly unlikely as

- GCE: bumped COS image version to `cos-beta-73-11647-64-0` ([#75149](https://github.com/kubernetes/kubernetes/pull/75149), [@yguo0905](https://github.com/yguo0905))

- Reworded log message upon image garbage collection failure to be more clear. ([#112631](https://github.com/kubernetes/kubernetes/pull/112631), [@tzneal](https://github.com/tzneal)) [SIG Node]
- Scheduler dumper now exposes a summary to indicate the number of pending pods in each internal queue. ([#111726](https://github.com/kubernetes/kubernetes/pull/111726), [@Huang-Wei](https://github.com/Huang-Wei)) [SIG Scheduling and Testing]