sue for patent infringement).  To "grant" such a patent license to a
party means to make such an agreement or commitment not to enforce a
patent against the party.

  If you convey a covered work, knowingly relying on a patent license,
and the Corresponding Source of the work is not available for anyone
to copy, free of charge and under the terms of this License, through a

        current release, GPU support for other floating-point types
        (`tf.float16`, `tf.float64`, `tf.complex64`, and `tf.complex128`) has
        been added for this op. If you were relying on the determinism of the
        `tf.float64` CPU implementation being automatically selected because of
        the absence of the `tf.float64` GPU implementation, you with either need

- Revised the logic for DaemonSet rolling update to exclude nodes if scheduling constraints are not met.

- `kubectl` no longer defaults to `http://localhost:8080`.  If you own one of these legacy clusters, you are *strongly- encouraged to secure your server.   If you cannot secure your server, you can set `KUBERNETES_MASTER` if you were relying on that behavior and you're a client-go user. Set `--server`, `--kubeconfig` or `KUBECONFIG` to make it work in `kubectl`. ([#86173](https://github.com/kubernetes/kubernetes/pull/86173), [@soltysh](https://github.com/soltysh)) [SIG API Machinery,...

- kubeadm now deletes the bootstrap-kubelet.conf file after TLS bootstrap
  User relying on bootstrap-kubelet.conf should switch to kubelet.conf that contains node credentials ([#80676](https://github.com/kubernetes/kubernetes/pull/80676), [@fabriziopandini](https://github.com/fabriziopandini))

Consider an application streaming a video over http/2. Perhaps the user pauses the video and the application stops reading bytes from this stream. The buffer will fill up, and flow control prevents the server from sending more data on this stream. When the user unpauses her video the buffer drains, the read is acknowledged, and the server proceeds to stream data.

## Retrying Requests

Sometimes connections fail: either a pooled connection was stale and disconnected, or the webserver itself couldn’t be reached. OkHttp will retry the request with a different route if one is available.

- Revised the logic for DaemonSet rolling update to exclude nodes if scheduling constraints are not met.

# Controllers

Istio has a variety of [controllers](https://kubernetes.io/docs/concepts/architecture/controller/), which basically watch some inputs and do something.
This can be reading from Kubernetes and writing other objects back, writing to proxies over XDS, etc.

Unfortunately, writing controllers is very error prone, even for seemingly simple cases.
To work around this, Istio has a variety of abstractions meant to make writing controllers easier.

## Clients

upgrade)

- A bug was fixed in kubelet where exec probe timeouts were not respected. This may result in unexpected behavior since the default timeout (if not specified) is `1s` which may be too small for some exec probes. Ensure that pods relying on this behavior are updated to correctly handle probe timeouts. See [configure probe](https://docs.k8s.io/tasks/configure-pod-container/configure-liveness-readiness-startup-probes/#configure-probes) section of the documentation for more details....