// OutCapacity is the output capacity on the caller.
		// If <= 0 capacity will be 1.
		OutCapacity int
	}

	// StreamHandlerFn must process a request with an optional initial payload.
	// It must keep consuming from 'in' until it returns.
	// 'in' and 'out' are independent.
	// The handler should never close out.
	// Buffers received from 'in'  can be recycled with PutByteBuffer.

    [Okio](https://github.com/square/okio), an I/O library that makes it easier
    to access, store and process data. Using this library internally makes OkHttp
    faster while consuming less memory. You can write a `RequestBody` as an Okio
    `BufferedSink` and a `ResponseBody` as an Okio `BufferedSource`. Standard
    `InputStream` and `OutputStream` access is also available.

application is **simple**, this will probably **not be a problem**, and you might not need to specify hard memory limits. But if you are **using a lot of memory** (for example with **machine learning** models), you should check how much memory you are consuming and adjust the **number of containers** that runs in **each machine** (and maybe add more machines to your cluster).

If you run **multiple processes per container** you will have to make sure that the number of processes started doesn't...

 *  Fix: When a connect times out, attempt an alternate route. Previously route
    selection was less efficient when differentiating failures.
 *  New: `Response.peekBody()` lets you access the response body without
    consuming it. This may be handy for interceptors!
 *  New: `HttpUrl.newBuilder()` resolves a link to a builder.
 *  New: Add the TLS version to the `Handshake`.
 *  New: Drop `Request.uri()` and `Request#urlString()`. Just use

- `k8s.io/kubernetes` and published components (such as `k8s.io/client-go` and `k8s.io/api`) now contain go module files including dependency version information. See [go-modules](http://git.k8s.io/client-go/INSTALL.md#go-modules) for details on consuming `k8s.io/client-go` using go modules. ([#74877](https://github.com/kubernetes/kubernetes/pull/74877), [@liggitt](https://github.com/liggitt))

#### Apps

### Autoscaling

 - Consuming metrics directly from Heapster is deprecated in favor of
   consuming metrics via an aggregated version of the resource metrics API.

   - In version 1.8, enable this behavior by setting the
     `--horizontal-pod-autoscaler-use-rest-clients` flag to `true`.

- Add Kubernetes Events to the `Kubelet Graceful Shutdown` feature. ([#1010...

- Fix: skip instance not found when decoupling vmss from lb ([#105836](https...

    * This solves a vulnerability that could allow denial of service attacks by using many small multipart fields/files (parts), consuming high CPU and memory.
    * Only applications using forms (e.g. file uploads) could be affected.
    * For most cases, upgrading won't have any breaking changes.

## 0.91.0

### Upgrades