type MonitorReaderOptions struct {
	BucketOptions
	HeaderSize int
}

// Read implements a throttled read
func (r *MonitoredReader) Read(buf []byte) (n int, err error) {
	if r.throttle == nil {
		return r.r.Read(buf)
	}
	if r.lastErr != nil {
		err = r.lastErr
		return
	}
	b := r.throttle.Burst()  // maximum available tokens
	need := len(buf)         // number of bytes requested by caller

	limitBytes := limit / int64(m.NodeCount)
	throttle, ok := m.bucketsThrottle[BucketOptions{Name: bucket, ReplicationARN: arn}]
	if !ok {
		throttle = &bucketThrottle{}
	}
	throttle.NodeBandwidthPerSec = limitBytes
	throttle.Limiter = rate.NewLimiter(rate.Limit(float64(limitBytes)), int(limitBytes))
	m.bucketsThrottle[BucketOptions{Name: bucket, ReplicationARN: arn}] = throttle
}

MinIO server allows to throttle incoming requests:

- limit the number of active requests allowed across the cluster
- limit the wait duration for each request in the queue

These values are enabled using server's configuration or environment variables.

      fail<Any>()
    } catch (expected: IOException) {
    }
    server.url("/b").toUrl().openConnection().getInputStream() // Should succeed.
  }

  /**
   * Throttle the request body by sleeping 500ms after every 3 bytes. With a 6-byte request, this
   * should yield one sleep for a total delay of 500ms.
   */
  @Test
  fun throttleRequest() {
    assumeNotWindows()

Framing rules make it impractical to implement http/2 correctly on a single blocking thread. The flow-control features introduce feedback between reads and writes, requiring writes to acknowledge reads and reads to throttle writes.

In OkHttp we expose a blocking API over a framed protocol. This document explains the code and policy that makes that work.

### Threads

#### Application's calling thread

    server.enqueue(MockResponse(body = "B"))
    val inputStream = server.url("/a").toUrl().openConnection().getInputStream()
    assertThat(inputStream!!.read()).isEqualTo('B'.code)
  }

  /**
   * Throttle the request body by sleeping 500ms after every 3 bytes. With a 6-byte request, this
   * should yield one sleep for a total delay of 500ms.
   */
  @Test
  fun throttleRequest() {
    assumeNotWindows()

   * allow in normal conditions. This is enforced by "throttling" incoming requests as needed. For
   * example, we could compute the appropriate throttle time for an incoming request, and make the
   * calling thread wait for that time.
   *
   * The simplest way to maintain a rate of QPS is to keep the timestamp of the last granted

```
KEY:
api  manage global HTTP API call specific features, such as throttling, authentication types, etc.

ARGS:

// the name displayed in logs and trace to be accurate. The name is extracted
// via reflection.
//
// When **no** flags are passed, the behavior is to trace both headers and body,
// gzip the response and throttle the handler via `maxClients`. Each of these
// can be disabled via the corresponding `s3HFlag`.
//
// CAUTION: for requests involving large req/resp bodies ensure to pass the

    assertThat(server.takeRequest().sequenceNumber).isEqualTo(1)
  }

  /**
   * Test to ensure we don't  throw a read timeout on responses that are progressing.  For this
   * case, we take a 4KiB body and throttle it to 1KiB/second.  We set the read timeout to two
   * seconds.  If our implementation is acting correctly, it will not throw, as it is progressing.
   */
  @ParameterizedTest