- Packets considered INVALID by conntrack are now dropped. In particular, this fixes
a problem where spurious retransmits in a long-running TCP connection to a service
IP could result in the connection being closed with the error "Connection reset by
peer" ([#74840](https://github.com/kubernetes/kubernetes/pull/74840), [@anfernee](https://github.com/anfernee))

    val response1 = client.newCall(request).execute()
    val source = response1.body.source()
    assertThat(source.readUtf8()).isEqualTo("ABC")

    // OpenJDK 6 fails on this line, complaining that the connection isn't open yet
    val cipherSuite = response1.handshake!!.cipherSuite
    val localCerts = response1.handshake!!.localCertificates
    val serverCerts = response1.handshake!!.peerCertificates

* Retry 'connection refused' errors when setting up clusters on GCE. ([#57394](https://github.com/kubernetes/kubernetes/pull/57394), [@mborsz](https://github.com/mborsz))
* Retry 'connection refused' errors when setting up clusters on GCE. ([#57394](https://github.com/kubernetes/kubernetes/pull/57394), [@mborsz](https://github.com/mborsz))

an L7 load balancer that already mitigates these http/2 attacks may choose not to enable the kube-apiserver mitigation to avoid disrupting load balancer → kube-apiserver connections if http/2 requests from multiple clients share the same backend connection. An API server on a private network may choose not to enable the kube-apiserver mitigation to prevent performance regressions for unauthenticated clients. Authenticated requests rely on the fix in golang.org/x/net v0.17.0 alone. https://issue.k8s.io/121197...

  -  Remove already deprecated options `resyncperiod` and `upstream` in the Kubernetes plugin.
  -  Includes Prometheus metrics name changes (to bring them in line with standard Prometheus metrics naming convention). They will be backward incompatible with existing reporting formulas that use the old metrics' names.
  -  The federation plugin (allows for v1 Kubernetes federation) has been removed.

an L7 load balancer that already mitigates these http/2 attacks may choose not to enable the kube-apiserver mitigation to avoid disrupting load balancer → kube-apiserver connections if http/2 requests from multiple clients share the same backend connection. An API server on a private network may choose not to enable the kube-apiserver mitigation to prevent performance regressions for unauthenticated clients. Authenticated requests rely on the fix in golang.org/x/net v0.17.0 alone. https://issue.k8s.io/121197...

- Kubernetes is now built with Go 1.19.9 ([#117776](https://github.com/kubernetes/kubernetes/pull/117776), [@xmudrii](https://github.com/xmudrii)) [SIG Release and Testing]

### Bug or Regression

- Fix "dbus: connection closed by user" error after dbus daemon restart
  CVE-2023-27561 CVE-2023-25809 CVE-2023-28642: Bump fix runc v1.1.4 -> v1.1.5

- Fixes a 1.30+ regression in connection stability for exec / attach / portforward requests initiated using a websocket client ([#130253](https://github.com/kubernetes/kubernetes/pull/130253), [@fuweid](https://github.com/fuweid)) [SIG API Machinery, CLI and Testing]

- Kube-proxy: fix a bug on UDP `NodePort` Services where stale connection tracking entries may blackhole the traffic directed to the `NodePort` ([#98305](https://github.com/kubernetes/kubernetes/pull/98305), [@aojea](https://github.com/aojea))

- Fixes a 1.30+ regression in connection stability for exec / attach / portforward requests initiated using a websocket client ([#130251](https://github.com/kubernetes/kubernetes/pull/130251), [@fuweid](https://github.com/fuweid)) [SIG API Machinery, CLI and Testing]