restarting the daemon alleviates this issue, but hasn’t been verified to
completely eliminate the IP allocation issue. More information [here](https://github.com/kubernetes/kubernetes/issues/21523#issuecomment-191498969)
  * Daemon becomes unresponsive (rarely) due to kernel deadlocks. More information [here](https://github.com/kubernetes/kubernetes/issues/21866#issuecomment-189492391)

### Provider-specific Notes

#### Various

   Core changes:

  final long expireAfterAccessNanos;

  /** How long after the last write to an entry the map will retain that entry. */
  final long expireAfterWriteNanos;

  /** How long after the last write an entry becomes a candidate for refresh. */
  final long refreshNanos;

  /** Entries waiting to be consumed by the removal listener. */
  // TODO(fry): define a new type which creates event objects and automates the clear logic

			// Experiments (using IndexPeriodic) suggest
			// the cutover is about 16 byte skips.
			// TODO: if large prefixes of sep are matching
			// we should cutover at even larger average skips,
			// because Equal becomes that much more expensive.
			// This code does not take that effect into account.
			j := bytealg.IndexRabinKarp(s[i:], sep)
			if j < 0 {
				return -1
			}
			return i + j
		}
	}
	return -1

### API Change

- A new alpha feature gate, `MutableCSINodeAllocatableCount`, has been introduced.
  

    Thread.sleep(TimeUnit.NANOSECONDS.toMillis(Http2Connection.DEGRADED_PONG_TIMEOUT_NS.toLong()))
    assertThat(connection.isHealthy(System.nanoTime())).isFalse()

    // When a pong does arrive, the connection becomes healthy again.
    connection.writePingAndAwaitPong()
    assertThat(connection.isHealthy(System.nanoTime())).isTrue()

    // Verify the peer received what was expected.

### DELETE operation in REST API

    * This ensures that attachment and detachment of volumes is independent of any single nodes’ availability. Meaning, if a node or kubelet becomes unavailable for any reason, the volumes attached to that node will be detached so they are free to be attached to other nodes.

		// the content implicitly for all versions, we can
		// avoid triggering another MRF heal for offline drives.
		if len(versions) == 0 {
			// Whether a disk was initially or becomes offline
			// during this upload, send it to the MRF list.
			for i := range len(onlineDisks) {
				if onlineDisks[i] != nil && onlineDisks[i].IsOnline() {
					continue
				}

    // 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
    // CONTAINS_VALUE_RETRIES increases, though without locking it is theoretically possible.
    Segment<K, V, E, S>[] segments = this.segments;
    long last = -1L;

  - kube_apiserver_pod_logs_pods_logs_backend_tls_failure_total becomes kube_apiserver_pod_logs_backend_tls_failure_total
  - kube_apiserver_pod_logs_pods_logs_insecure_backend_total becomes kube_apiserver_pod_logs_insecure_backend_total ([#114497](https://github.com/kubernetes/kubernetes/pull/114497), [@dgrisonnet](https://github.com/dgrisonnet))